2026-06-07 - SSN - Payrolls Related Selloff, Memory Beating, Iran War
Update, Gold Beating Treasuries as Reserve, Special Sits Update MGM,
SpaceX IPO, PFE Value 6.5Pct Divvy (transcript)

Alright, happy Sunday everyone. Um, last Friday we had a very jarring sell-off in the market, uh, given payrolls was much,
much stronger than anyone had expected. Um, even though a lot of that was temporary lodging uh, hires ahead of the the
summer vacation season.

We also had a 15% sell-off in DRAM. Um, you know, memory and Korean stocks had gone up in a parabolic line. So we had
a pullback, um, because of that. Um, after the payrolls report, you know, the 10-year was up several basis points on Friday,
um, resulting in interest rate fears, uh, which were exacerbated by a lack of a ceasefire and, um, you know, a lack of a
peace treaty between Iran after, you know, we we've been in this war for about 100 days now. Um, the war that's not a
war apparently.

We'll also talk about how the ECB said last week that gold has overtaken US Treasuries as the top reserve asset and what
that means. We will also discuss, you know, other reasons why the Nasdaq was down 4% on Friday, including, you know,
SpaceX, uh, taking up new capital, um, you know, the IPO risks associated, um, with this upcoming IPO. Um, we'll talk
about the Korean stock bubble. We'll talk about Google um, issuing 80 billion of stock after years of of buying back stock
and rumors that Meta is going to issue stock. Um, so obviously several moving parts and then on the merger arb and
special sits side, um, you know, we will, um, we will discuss

MGM, uh, which was bid, um, on last week just like Caesars was bid on. You know, they have the BetMGM assets, uh,
both Vegas and uh, Macau assets. Barry Diller, um, you know, IAC's been renamed to Peoples, um, they effectively bid
48.30 for the business. We'll talk about, um, Berkshire Hathaway, uh, given Abel is running the show, uh, and Warren
Buffett's retired, you know, their acquisition of TMHC, and we'll give an update on uh, Warner Brothers WBD as well.

PDF Overview and Pfizer Model - 02:27

So, you know, if you if you open your PDF, um, which is saved both in the SSR Q&A tab, um, and should be on your email
as well, um, you know, you'll have a model of Pfizer, uh, which we think is worth $32 a share, um, both on a probability-
weighted basis, um, and on an intrinsic DCF basis. Uh, you'll also see, um, you know, some interesting PDFs, you know,
from Merrill, Citi, Standard Chartered on the market, IPOs, inflation risks, earnings.

Um, and if you flip to page one of the PDF, um, it starts with, you know, the upcoming economic calendar, um, earnings
and the merger arb update for the past week. Um, on page six, uh, we talk about how gold has overtaken US Treasuries as
a top reserve asset. On page 12, we talk about the market correction and and five reasons for the volatility. On page 16,
we talk about the Iran war and upside risks in oil. On page 18, we talk about, you know, how if you remove AI and you
know, from the SPX, uh, the rally effectively disappears since the start of the Iran war on February 28th. We then discuss
briefly the SpaceX IPO discussion, Demodaran interviews, um, where the IPO is trading in Hyperliquid, which is uh far
below the IPO price at 135, it's trading like around 170.

Uh, we'll talk about the memory stock sell-off, uh, how DRAM fell 15% and, you know, Korean stocks were down 6% on
Friday. We'll talk about software stocks, uh, rallying, uh, on Jensen Huang's conciliatory tone, on Jensen saying Marvell is
the next $1 trillion company after buying 2 billion of stock, uh, which then sold off on Friday. We'll talk about the CPU
shortage discussion, um, which is a little bit different than the GPU shortage discussion. Uh, you know, give the bullish
perspective from uh, Tang AI. On page 38, we go through MS's takeaways uh from meetings in Taiwan on semiconductors,
which were upbeat. On page 42, we talk about data center construction spending, uh, and how that eclipses public sector
infrastructure spend now. Page 44, we'll discuss um Meta's megawatt capacity coming online, a recent hack and insider
selling. Page 46, we'll discuss uh HPE being up 25%, which outperformed like Dell last week. Um, page 47, we talk about
the AI labor debate and how Yann LeCun comes out ahead, uh, for now saying fears for labor are overblown. Page 49, we
discuss how AI can go wrong, how a pizza franchisee sued the parent company for 100 million, alleging the AI delivery
system caused huge operational breakdowns and delivery delays. Page 51, we'll discuss Google's capital raise of 84 billion.
And page 52, we'll discuss the American savings rate. And finally on page 53, we'll we'll discuss Pfizer's uh valuation
model. So lots to a lot to go over here.
Upcoming Economic Calendar - 05:35

Um, with respect to the economic calendar for the coming week, um, New York Fed expectations next week, see, will
likely see an increase in inflation expectations, um, from 3.6% to 3.7%. NFIB, uh, small business optimism is expected to
be flat. Uh, trade balance, negative 56 billion on the ninth. Also on the ninth, we'll have existing home sales, which are
expected to be flat. Um, CPI is the most important data point, um, for next week. Um, you know, just given inflation
expectations, um, after CPI, we have um, initial jobless claims on the 11th and then PPI, um, on the 11th as well. So we
have, it's a double whammy for inflation with CPI and PPI both printing next week. Um, so, I mean, the 10-year is already at
4.54, um, which is above kind of Besant's danger, danger zone. Now, if we do get a hot PPI or CPI, um, I do think that could
be a real problem uh, for the market next week. So, while we should get a bounce outside of, um, inflation spiking, um,
and outside of a strong escalation in the war after what we saw on Friday, um, I do think that, um, I do think that, you
know, if you look at over markets over the weekend, you are seeing a small rally in Bitcoin to 63,000. You know, it was
below 60,000 for a moment. You're seeing Ethereum at 1674. Um, you know, if you look at, um, futures right now, you are
seeing, um, you know, a slightly negative print, you know, down 10 bips on the S&P 500 because of the slight war
escalation over the weekend, small caps down 30 bips. Um, but you're not seeing that cascade lower that we saw, um, on
Friday. But again, you know, next week, the CPI, PPI and then UMich inflation expectations, you know, which are supposed
to be for one year, 4.9% and the five to 10 year, which is now 4%, you know, the biggest risk, um, is global rates, um, in my
opinion.

Upcoming Earnings Reports - 08:17

On the following page, you know, earning season is basically over, but on Monday, we do have Campbell's, uh,
Campbell's reporting pre-market, Duluth Trading, um, Vail Resorts after the close on Monday. On Tuesday, we have
Academy Sports, um, Lands' End, um, after the close, you have, uh, BarkBox, Cracker Barrel, Skillsoft. On Wednesday,
you have Chewy, and after the close, you have Oracle, which is probably the most important earnings of the week. Um,
and then finally on Thursday, you have Adobe on the software side, Lennar on the home builder side, and McGraw Hill.

Merger and Special Situations Update - 08:54

In terms of the the special situations, merger arb update, MGM Resorts, um, you know, got a bid, 48.30 from IAC last
week. On Friday, shares closed down 4.5%, finishing essentially in line with the proposal. They're actually trading through
the proposal, potentially expecting a, um, bidding war. You know, if shares fall again tomorrow below 48, you know, we
might consider starting a position there. We didn't start a position when the deal was announced because shares, you
know, automatically rallied higher than the bid, um, which was an all-cash bid. So, um, I didn't feel comfortable with that.

Assuming a similar deal spread as Caesars and a deal completion in about one year, the trading price would imply $52 a
share deal price, um, is expected. So the market's expecting a bump because there should be a spread. Okay? Uh, the deal
price has been the primary debate, uh, in our conversations with most views ranging from low 50s to mid 50s, but some
valuing the business at 60% plus based on the sum of the parts approach around MGM's various assets, which we'll talk
about in a moment. As a 50% stake, uh, in BetMGM, the other 50% is owned by Entain. It's MGM Osaka options in Japan,
uh, the 56% stake in MGM China as well. To this point, MGM CFO acknowledged the complexity in valuing the business,
indicating sometimes investors just aren't doing the work to value the sum of the parts and highlighting the low implied
valuation on its US business.

Aside from valuation, other focus items include, number one, financing details. IAC needs 3 billion to get to 50.1% and
additional equity partners for 6 billion. Investors also focus on certain leverage covenants on the existing assets. Two,
IAC's motivation, deal or put assets in play, although investors are hard-pressed to identify another likely buyer. And
three, IAC's longer-term plan for Entain and BetMGM.

In other news, Honeywell Aero hosted its investor day three weeks ahead of the spin. You can probably find the
presentation online. I'll I'll look for it as well after this call. ProAssurance reiterated its closing guidance by June 30th, 2026,
and the EC has commenced its formal review of the Warner Brothers Pies-guy transaction with a phase one deadline for
July 7th. And so we bought a little bit of that around a 14% spread, which you see is number two on, um, you know, the
most attractive spreads in the large cap space.
Um, recent transactions on June 1st, um, outside of MGM, you had the Berkshire TMHC, uh, all cash offer which trades at
a tiny, tiny spread because it's Berkshire. Um, at the end of last week, you had the AKZA, the Nippon Paint and Sherwin-
Williams, um, ending their pursuit of Akzo Nobel. Nippon Paint Holdings and Sherwin-Williams ended their effort to
buy Akzo Nobel, leaving the Dutch paint maker to proceed with its planned merger with US firm Exalta. The decision
follows Akzo Nobel's rejection of two joint all-cash offers put forward by the firms. The companies, um, said the most
recent offer of 73 euros per share, um, valued the maker of Dulux paints at about 12.5 billion euros or 14.5 billion dollars,
but it involved plans to break up the business, which the company's management didn't want.

Now, the current merger dynamics around MGM Resorts International are being driven by three overlapping themes.
One, consolidation in US gaming. Two, the strategic value of BetMGM, and three, the belief that MGM is undervalued
relative to its real estate and digital optionality. The most important recent development in Barry Diller's Peoples, Inc.
takeover of, uh, proposal for MGM, which People, Inc. is the former IAC, which values MGM at roughly 18 billion or 48.30,
uh, per share in cash is, um, you know, well, first, let's talk about People. People, Inc., uh, or formerly IAC, already owns
about 26.1% of MGM. It's basically trying to acquire the remaining stake.

So, MGM controls roughly 40% of the Las Vegas Strip through properties like The Bellagio, Aria, MGM Grand,
Cosmopolitan operations, Mandalay Bay. Um, those assets are difficult to replicate and generate large recurring cash
flows. Diller specifically described them as real-world assets that AI cannot easily replicate. On BetMGM, the digital betting
JV with Entain is increasingly central to the M&A story. BetMGM has inflected from cash burn into profitability. 2025
revenue of about 2.8 billion, EBITDA around, uh, 220 million. Um, positive quarterly cash flow generation, top-tier US
iGaming position. Historically, investors viewed, um, BetMGM as a drag because of marketing losses. Now it has become a
valuable strategic asset.

The real question is whether MGM eventually buys one, all of Entain, or just Entain's 50% stake in BetMGM. This has
been rumored repeatedly for years. Um, BetMGM is currently owned 50% by MGM, 50% by Entain, and the problem is
that MGM controls, um, physical casinos and a US customer funnel. Entain provides much of the sportsbook technology
and international gaming expertise, and that split limits its strategic flexibility. Analysts increasingly believe full ownership
would one, simplify capital allocation, two, allow easier IPO spin-off options, three, improve economics, and four,
potentially unlock a higher multiple.

Now, the Macau and international exposure that MGM owns is probably also being undervalued here, um, because of the
political issues between the US and and Asia right now and lower multiples in Asia. MGM China gives exposure to the
Macau recovery and Asian gaming demand. Diller specifically highlighted MGM China's exposure as under-appreciated.
Um, MGM's further discussion. So, MGM CFO Jonathan Halkyard states, "Investors aren't doing the work to value the
company properly." Uh, MGM Resorts' Chief Financial Officer Jonathan Halkyard echoed Barry Diller's sentiment that
markets aren't giving the casino giant its due. Speaking at a Hotel Investment Conference in New York on Tuesday,
Halkyard called Diller's assessment, um, of MGM gratifying a day after the magnet's People, Inc. offered to take the
company private in a 12.4 billion deal. People, Inc. currently holds a 26% stake in MGM.

"Sometimes investors just aren't doing the work to value the sum of the parts," Halkyard said during a discussion of future
gaming hospitality. He suggested that the complexity of MGM's business made it more difficult to assess. He also noted
that the value of the company's domestic business, which includes operations in Las Vegas, was at a very low multiple. He
said that MGM could improve how it was communicating its value. Initially, shares of MGM were up 16% on June 1st,
following People's all-cash proposal. We think that, you know, below, um, you know, below 45, um, this is probably a buy.
Um, because there will probably be a topping bid and it will likely rally on the topping bid before selling off again.

You know, as a benchmark, if you look at Fertitta's recent acquisition of Caesar, a couple data points, the premium, you
know, was roughly a 49% premium to its February 25th share price. Applying the same premium at the same point to
MGM, which was at an unaffected price of 35.56 on 2/25, uh, would imply a 52.98, uh, share takeout price for MGM. This
deal represents about a 6.6 times multiple on 2027 EBITDA on consensus.

Now with respect to Warner Brothers, uh, the EC commences a formal review on the Pies-guy transaction. It's a 14%
spread and the phase one deadline is set for July 7th. So we do think that the the Pies-guy spread is quite attractive. Um,
you guys can read about PRA and Estée Lauder CEO, um, who spoke about, uh, despite failed talks with with, um, you
know, the company was trying to acquire, uh, Puig, um, it's still open to acquisitions if they make financial sense. The
reason why they broke the deal was, um, because the stock sold off so dramatically after they announced it, right? You
know, French companies just trade a very big discounts as well, um, you know, for regulatory and, uh, you know, just
difficulty in firing people, just high taxes, etc.

Berkshire Hathaway also last week said they were going to acquire Taylor Morrison Home Corporation for 8.5 billion.
Taylor Morrison Home Corporation, um, you know, was bought out at about $72.50 per common share in cash,
representing 6.8 billion equity and and total enterprise value of 8.5 billion. It was a 24% premium to Taylor Morrison's
latest closing price, and because it's Berkshire, they just expect accepted the deal and the spread is super tight, 1.4%. Um,
Universal Music Group rejected the $65 billion buyout offer again from from Bill Ackman. Um, so largest music company
in the world after Universal's largest shareholder voiced opposition. They said that the value of of the offer fundamentally
and materially undervalued UMG. So they're asking for a higher bid.

Notable spread changes, you know, at the end of May, you can kind of see what tightened and what what widened. You
know, there's obviously that that Take-Two uh bidding war. Um, UNF rallied on FTC optimism, uh, Avego earnings were
decent, uh, Quervo, uh, slight tightening, um, and then the deals that underperformed were PRA and A.C.

ECB Report: Gold Overtakes US Treasuries as Top Reserve Asset - 19:31

ECB has stated that gold has overtaken US Treasuries as the top reserve asset. This is a very big, uh, a very important, um,
development. Um, you know, this is a landmark shift. The European Central Bank confirmed that gold has officially taken
US government bonds to become the world's largest asset. You can see that in a graph by Tavi Costa, first time since 1996.
The changing dynamics behind this shift in which is why we think gold is a buy after this sell-off along with, uh, even silver.
You know, by the numbers, by the end of 2025, gold accounted for about 27% of total official foreign reserves, pushing it
just ahead of US Treasuries, um, in the low 20% range. Central banks now hold close to about 4 trillion in gold. ECB
President Christine Lagarde cited geopolitical tensions and an increase in reserve diversification as primary catalysts.
Following the weaponization of the US and allied sanctions such as the freezing of Russian reserves in 2022, many nations,
particularly emerging market central banks like Poland, India, Turkey have been aggressively acquiring gold to avoid
counterparty and political risks.

Global Central Banks Sell US Treasuries - 20:36

Um, with respect to the BRICS, you have China, the People's Bank of China has added hundreds of tons of gold while
simultaneously allowing its holdings of US Treasuries to drop significantly. In India, the Reserve Bank of India has
expanded its gold reserves to literally 900 tons, prioritizing bullion to protect the rupee against global monetary shifts. In
Russia, following the freeze of its 300 billion in Western forex reserves, Russia completely pivoted to gold and domestic
currencies, actively mining and hoarding gold to bypass the Swift dollar network. Brazil, breaking a multi-year hiatus, Brazil
re-entered the gold market as a net buyer, aggressively adding to its gold reserves. In Europe, you had Poland, the
Narodowy Bank Polski is the most aggressive Western buyer. It increased its gold holdings to 582 tons with an official
institutional target of reaching 700 tons to insulate itself from regional instability. Turkey, navigating severe domestic
inflation, Turkey's Central Bank has consistently ranked among the top net buyers, heavily expanding its gold stock pile at
the direct expense of fiat currency reserves. But the country had to recently trim its gold reserves to protect its currency,
given that it's an oil importer during the Iran war. With respect to Austria and Hungary, both Central European nations
have notably increased their structural share of gold following Poland on their balance sheets to hedge against currency
shocks.

With respect to the price rally, the headline figure is partly structural and partly a result of market mechanics. Aggressive
central bank buying compounded with historical, uh, with a historical gold boom market, um, that was influenced by China
retail buying, saw the metal climb to record highs above 4,500 an ounce, artificially inflating the share of gold on central
bank balance sheets as well. Broader implications, while this signals a quiet vote of no confidence in net in the neutral
utility of dollar-based financial system, the US dollar still remains the most dominant transactional and invoicing currency
globally. However, it has fundamentally weakened the structural buyer base, um, for US long-term debt and the the net
biggest buyer going forward will have to be US citizens and banks and institutions.
So you can see a summary of the central bank reserve shift, uh, in a table below and countries selling Treasuries are on
page seven. So global foreign holdings of US Treasury bonds have faced intense structural pressure, while total
outstanding debt has scaled to new heights. Specific emerging markets and geopolitical actors have aggressively pared
back their allocations to limit vulnerability to the dollar ecosystem. With respect to China, the drawdown under the
direction of the People's Bank of China, Beijing has reduced its US Treasury holdings by roughly 50% from its historical
peak of 1.3 trillion. In terms of recent activity, holdings fell sharply to 652.3 billion down from 700 billion just in a single
month. Regulators prompted domestic banks to aggressively sell 41 billion in paper. Funds are being explicitly reallocated
into physical gold billion and harder assets to insulate from potential sanction vulnerabilities.

In Japan, while Japan remains the single largest foreign holder of US debt, its baseline holdings contracted by over 5%,
driven by aggressive domestic currency interventions to stabilize the Yen. Tokyo liquidated a staggering 47.7 billion in US
Treasuries recently. The this massive exit underscores a broader trend, using US paper debt as liquid capital to protect
domestic fiscal baselines rather than as a permanent store of value. In India, um, core emerging and other emerging
markets, core emerging market economies slash their combined US Treasury positions by 108 billion over a 12-month
period. Um, even the Reserve Bank of India, you know, India being a uh, an ally, reduced its holdings, um, by 7.6 billion
recently. Proceeds were challenged into, challenged, channeled into expanding India's sovereign gold stock pile to nearly
900 tons, building domestic gold's share of total reserves to 16.7%.

Now in the Middle East, um, Gulf states have actively stepped away from the traditional rigid reliance on the petrodollar
recycling system. In some of you even adopted a petro-Yuan. Saudi Arabia liquidated 10.8 billion of US Treasuries and the
UAE sold 5.8 billion in US Treasuries recently. Um, sovereign wealth funds are diversifying trade surpluses directly into
alternative currencies and physical gold infrastructure, and you can see a table of all this movement with central bank gold
reserve changes, the biggest changes from 2020 to 2025, uh, in the table below with China, Poland, Turkey, India, Brazil,
um, being the biggest buyers, followed by Azerbaijan, Japan, Thailand, Hungary, Singapore, Iraq, Qatar, Russia, and the
UAE.

Repatriation of Gold Reserves - 25:17

There's also been a big move of repatriation of gold reserve assets. So according to the World Gold Council Central Bank
Survey, 59% of of global central banks now opt to store their gold reserves domestically up sharply from 41%. This shift
marks the decline of the post-World War II American vault illusion where foreign central banks traditionally left physical
metal under the custody of the US Federal Reserve. In France, the Banque de France executed a massive discreet swap
operation to pull 129 tons of gold out of the New York Fed. Instead of shipping gold, uh, ingots across the Atlantic, France
basically sold its legacy bars into New York and simultaneously pre-purchased modern London Good Delivery bars to be
stored securely in Paris. The result is that France now holds 100% of its 2,437 ton reserve on domestic soil for the first time
since the interwar period, capturing a windfall in billions in capital gains due to the metal's price spike. In India, the
Reserve Bank systematically executed one of the largest physical movements of bullion in recent monetary history,
moving away from decades of relying on Western custodians, India physically shifted hundreds of tons from vaults in
London and Europe back to local depositories. India's domestic gold holdings expanded drastically, surging from 301 tons
to 680 tons in a period of several months. In Germany, the Deutsche Bundesbank is facing immense domestic political and
institutional pressure to bring its gold home from the United States. Critics pointed to a transatlantic trade friction and
shifting US administration's uh views as a structural risk to leaving sovereign assets exposed abroad. As a result, um,
there's 1,236 metric tons of German gold still sitting in New York representing more than a third of Germany's total
sovereign supply. In Poland and Serbia, you saw smaller, highly active Eastern European central banks completely rewiring
their custody frameworks. Serbia mandated that 100% of its national gold must be held within state lines. Poland also
aggressively expanded its baseline holdings toward a 700 ton target, insisting on direct domestic control to protect against
regional geoeconomic shocks.

Implications of the Shift from US Treasuries - 27:33

When central banks reduce their allocations to US Treasuries, they diminish the pool of price-insensitive long-term buyers.
This dynamic creates specific pressures along the US yield curve. So rising long-term yields as central banks sell off long-
term notes such as the 10 and 30-year bond, bond prices drop because bond prices and yields move in opposite
directions. This lack of structural demand forces 10-year treasury yields to be elevated, you know, now above 4.5%, you
know, on top of the recent commodity spike and the better than expected jobs report. The term premium spikes,
investors now demand higher interest payments to hold US long-term debt because foreign states no longer view
Treasuries as completely risk-free, the government must pay a higher risk premium to attract alternative capital. And
that's because of our $3 trillion deficit and, you know, half a trillion of additional military spending expected next year. The
constraining of the Federal Reserve, the exit of foreign sovereign buyers complicates monetary policy, even as markets
look for interest rate relief, stickier inflation and the need to keep US debt attractive mean that the Fed has been forced to
hold its benchmark rate steady in the 3.5 to 3.75 range with little room for aggressive cuts.

Now, surging US debt servicing costs, the combination of higher interest rates and a ballooning national debt means that
maintaining US obligations cost the government over 734 billion annually as of 2024. By 2026, it could be $1 trillion in
interest costs, roughly 17% of total federal spending and it will only rise from here. The latest projected annual interest
cost for US federal debt in 2026 is about 1.037 trillion according to J.P. Morgan. The GDP impact here is about 3.5% of
GDP. The annual interest cost has officially become one of the largest line items in the federal budget and now surpasses
total annual US defense spending, which is actually supposed to go up further, which will increase interest expense further
next year. This figure represents a near tripling of the federal debt servicing cost in just six years, climbing from only 345
billion in interest to about a trillion now. High net interest payments are compounding with rising mandatory spending on
entitlements like social security and major medical programs to push the projected 2026 federal budget deficit to about
1.9 trillion before the increase in in defense spending. In terms of the long-term trajectory, according to the latest 10-year
projections from the Congressional Budget Office (CBO), if borrowing trajectories continue on their current path, annual
US interest payments are projected to balloon over 2 trillion by the year 2036, or well over 30% of our our budget. You can
look at federal interest costs as a percentage of GDP, we're at, uh, highs, you know, not seen since, uh, the 1980s. Average
interest expense on the US government debt is now over 3 billion per day. By the end of the next three decades, interest
will consume about 7% of GDP and 35% of federal revenues annually. High and rising national debt will mean that more of
the budget will go towards servicing debt with interest payments instead of going towards other priorities. Importantly, a
high interest burden also makes it more difficult for lawmakers to borrow more in times of emergency or during a war
without significant consequences. Policymakers should work together to address these fiscal challenges before it's too
late. You can see spending on interest to exceed all other programs on CRFP.org and then gold having overtaken US dollar
on page 11.

Market Correction Analysis - 30:59

Now, let's talk about the market correction last week. We'll do a like a speed commentary on the five reasons for volatility,
but, you know, after we go through this overview. So we saw risk off into the weekend with the NDX, the Nasdaq down 477
bips or 4.8% and the SPX down 2.7%. On a hotter NFP print coupled with a follow through weakness post Broadcom
earnings and Anthropic's recursive learning headlines. Activity levels did not pick up despite velocity of the of the moves.
As expected, pockets that led the way lower were 12-month winners, drones, space, quantum computing, Bitcoin
sensitives and non-profitable tech all down over 10% on the day. SPX ex-AI actually finished unchanged on the day, believe
it or not, with Pfizer being up where we've shared our model. On the week, overall flow skews were benign at the client
level clear broadening out, uh, dynamic was present under the hood with megacap supply being a clear trend over the last
few sessions with Google and a number of big tech names selling off versus demand in healthcare. So MedTech, uh,
pharma, financials rallied.

Um, on the Goldman Prime Desk, hedge funds net bought US equities for a third week at the fastest pace in over six
months, driven by long buys outpacing short sales across both single stocks and macro products. Eight of 11 sectors were
net bought, led by industrials, which saw the largest net buying in five months, and Infotech, which has now seen de-
grossing activity in five of the last seven weeks. Um, at the same time, US hedge fund leverage fell 2.1 points to 208%. Um,
while US long-short net leverage fell 1.3 points to 53.9%.

Um, in terms of next week, the S&P implied move through next Friday is positive 1.6%. Uh, the market's expecting a
bounce. You have a busy week ahead with CPI on Wednesday, which we talked about, the main macro focus, following uh,
hotter NFP print this week. The FOMC is entering its blackout period ahead of rate decision on June 17th. While we get the
ECB rate decision on Thursday. On the micro side, watching a busy week of IPO issuance, a handful of sell-side
conferences like Goldman's healthcare conference on Monday to Wednesday, um, and then weekend headlines coming
from the OPEC meeting where OPEC is supposed to increase supply, which is more figurative than actual. Um, they're
trying to control the back end of the oil curve ahead of US driving season.

Non-farm payrolls rose 172k in May, well above expectations versus 85k consensus and a 92k whisper. In addition, there
were large upward two-month revisions and uh, plus of of plus 93k additional jobs, which was quite, quite impressive.
Unemployment was in line at about 4.3% unrounded. Market is now pricing close to one hike through year end with the
10-year back at 4.5%. Bottom line is the strong jobs numbers and stable unemployment rate increase the risk of a longer
Fed pause, though we still view rate hikes as unlikely this year. The Nasdaq closed out the week, uh, you know, down 430
basis points, erasing the past two weeks of gains, uh, and today and and Friday's one-day move was the worst for the
index since April of 2025. It was a frantic Friday of headline digestion amid a demanding technical backdrop. You know,
think of the semiconductor stocks, uh, closing 75% above its 200 DMA on Wednesday, the most extended above its 200
DMA since 1999-2000 era and then selling off. As rates outlook back, uh, front of mind after a strong NFP print pushed the
market to price in a full hike by the end of 2026, the semiconductor index then puked, down almost 10%.

Despite headline price action and sentiment, SPX, Nasdaq, and even SOX vol was actually down on the day across the
curve, uh, and Goldman's derivatives desk saw very little panic for hedges. That said, the resilience of the AI semi-trade
has been tested with SOX down by 9.7% on the day, 4.2% on the week, on the combination of Broadcom earnings miss,
Anthropic's recursive learning headlines, and increasing focus on equity market supply and issuance with upcoming IPOs
from SpaceX and OpenAI. Next week, the market will focus on WWDC and Oracle earnings and various software investor
days and user conferences, you know, think of Net, RBRK, DataDog, etc. The desk will be watching for positive AI data
points to arrest recent narrative and price action. After a nine-week winning streak at a breakneck pace, we are, you know,
SSR are not surprised to see some digestion of recent gains. We think we are seeing this volatility for five key reasons.

One, positioning for large-cap US equities, and specifically mega-cap tech, began became aggressively overweight at the
same time cash flows were flipping negative and buybacks were stopping. The chart below shows this breakdown from
Deutsche Bank, highlighting how large-cap tech positions are now in the 95th percentile, raising questions about who can
be the incremental buyers, mostly as we are seeing a surge in equity supply with $80 billion IPOs like SpaceX and OpenAI
and existing listing issuance from companies like Google.

Two, signs of aggressive risk-taking in options markets with investors clamoring for upside optionality and having little
demand for downside protection. And huge rallies in the lowest quality portions of the market are not great timing tools
for calling a near-term top, as we've written about frequently, silliness can persist. However, we do see these signs of
unbridled risk appetite, we see it as a sign of growing fragility under the surface. This high beta, low quality, high
momentum leadership can quickly unwind, which is exactly what we saw Friday, as seen in the chart below with the
Goldman Sachs unprofitable tech index.

Three, the valuation reset that drove the huge move in semiconductor stocks, which have been the key driver of the S&P
500 given their 92% rally off the March lows and 19% weight now in the SPX, has run its course. Memory stocks were
hammered since the SemiAnalysis report showed a system-level DRAM capacity per Vera Rubin rack is being configured at
only 28 terabytes, down from the previously estimated 55% terabytes, which means the new Vera Rubin Nvidia racks will
require half the memory. Semiconductor strength has certainly been driven by a huge earnings growth, but it's important
to appreciate that this earnings growth reset came prior to the strength in the stocks with the 92% rally off the March lows
coming from multiple expansion as valuation rebounded from recent lows. Effectively, the stocks got cheap during the
Iran War beginning as prices remained flat and earnings grew, but now the stocks are no longer cheap after the recent
vault higher. The chart below shows this dynamic with the SOX semiconductor PE ratio falling from 26 to 17 times earlier
in the year and then rebounding all the way to 28x, nearly the 2025 valuation peak. Friday's memory ETF, the DRAM, uh,
fell 15%, which is a move that was shocking to many.

Four, economic data is boxing the Fed into a less friendly and potentially more adversarial stance, sending yields higher as
the odds of a Fed rate hike increase, with the bond market now implying a 100% chance of a hike by December. Friday's
better-than-expected job state, a plus 172k jobs added versus 88 consensus with positive 93k two-month revisions and a
4.3% unemployment rate sent the two-year Treasury up 10 basis points on the day to a new high for 2026 and 40 basis
points above the current Fed funds rate as seen in the second chart below. You can see the labor print on page 14 and the
two-year rally on page 14.
Number five, market technicals were extended and losing steam at the same time. If you now look, the futures are actually
down about 30 bips on the S&P, down 30 bips on the Nasdaq, and down 50 bips um, on the, uh, Russell, and, you know, I
think that the market is just getting concerned about, you know, war headlines that are coming through overnight and
also selling in Asia. Anyway, number five, market technicals were extended and losing steam at the same time, meaning
we've moved very far, very fast to the upside, and there are brewing signs of exhaustion. We can measure this by looking
at overbought conditions with relative strength indices on both daily and weekly measures and comparing them to fading
momentum indicators like the MACD, which has been moderating since mid-May despite the new highs. The charts are
below on page 14.

The big question going forward will be whether or not the market will interpret this hawkish Fed as a potential drain on
liquidity, which we'll see with CPI and PPI next week. This could be a dampener of growth as well. The key metric to watch
will be credit spreads, which capture the bond market's potential consternation about growth. If corporate credit spreads
widen out materially, it could reflect expectations that this foe not friend Fed will weigh on growth. Note in the second
below, the widening has not yet occurred in credit. It's really just been a memory-led sell-off in equities. After such a
powerful run off the March lows, mostly one driven by a narrow cohort of high-beta, high-momentum stocks, we should
not be surprised that we are seeing a come-down from huge gains. Though many will push back against these qualms
about technicals, positioning, sentiment, valuation, and policy, citing ultra-strong earnings growth as a reason to ignore
these factors, as we showed over the last two weeks, there are nuances that investors should appreciate about the
forward earnings picture as well, such as the huge benefit from one-time gains like the Anthropic revaluations in 1Q26
and the largely unsustainable surge in semiconductor margins over the medium term. This creates an earnings backdrop
that remains robust in the near term but should be examined closely for deceleration by 2027. All of this is to say that we
do continue to see a bull market in stock prices and earnings driven by the earnings, but that we also see a healthy come-
down driven by technicals, positioning, sentiment, valuation, and policy when the market thinks about 2027 sustainability.

Other Sector Commentary - 41:23

Other sector commentary, you know, from Goldman, we we talk about the consumer. The consumer group traded heavy
in the early part of the week as it continues to trade, uh, the inverse of momentum. That is why we got further declines on
Friday, but not nearly as severe as the market as a whole. Within the consumer, the standout on the week was gaming,
lodging, and leisure names. The GS Gaming basket was up 4% on more M&A headlines like the MGM deal and better
fundamentals versus the retail market. More companies confirmed a slowdown for the start of 2Q, including Ulta Beauty
and Lululemon. Those are largely idiosyncratic, but investors are focusing more on those companies that have told us
things have slowed down, as opposed to noting trends that that have been resilient.

In terms of financials and banks, uh, the bank index outperformed the SPX by 4% in in the week that served as a reminder
that the sector remains mostly inversely correlated to tech and momentum. Upside was led by banks, money, money
center banks, followed by trusts, then regionals, with the large banks representing investors' favorite way to play bullish
capital market trends, and two, traditional asset managers, which alongside banks, now make up the top two clean longs
left in financials, while wealth and retail brokers saw a quiet relief rally as investor intention around terminal growth risks
shifted to the exchanges, which traded down, you know, 5 to 17% on the week with the CBOE, MIAX, uh, selling off as
investors grappled with the potential risk to retail trading volumes from perpetual futures following Coinbase and call
she's announcement of regulated perpetual futures for crypto on their platforms.

Last week also saw Nareit occur, an important, uh, REIT industry conference for real estate, where investor feedback came
away notably positive on lodging, plus 7% um, versus overall REITs down 1% on the week and residential, which was up 3%
on the week. Next week we will have the largest mid-quarter conference for financials coming up where investor inbounds
on the banks largely expect neutral to positive updates, though there is some vigilance from investors around any signals
of more intense deposit competition.

Energy was up 2% over the last five days, but down 3% over the month. Investors broadly remain bullish on the setup for
energy equities, um, where we continue to see more investors looking to buy the dip in energy equities on the back of
potential, uh, resolution, lack of resolution in the Middle East. Looking at subsectors, OFS sentiment remains broadly
strong across North America and international, despite softer price action in in, uh, oil services. There's been, um, more
recent bullish sentiment on refining as well. Natural gas feedback in the US continues to lead bearish, uh, temporarily, uh,
with Henry Hub plus specialist highlight relative bull cases for TTF into the summer and the fall.

Industrials were up 100 basis points last week, led by transports, while materials were down 60 bips. Both outperformed
the broader markets, which were down 150 bips, led to on the downside by tech. Upside was led by TMHC, which agreed
to be bought by Berkshire Hathaway to start the week at a 24% premium. SAIC was up 10% on the week with earnings
that came in well ahead of expectations and raising full-year EBITDA and EPS guidance, just a quarter into their fiscal year.
ODFL was up 8% on the week after monthly tonnage data and pricing came ahead of expectations. JCI was up 8% on the
week after raising topline organic growth guidance along with the Gemba investor day. On the downside, more
speculative names like Rocket Lab, FLY, KRMN, AVAV were down 10 to 20%. Ford reversed recent strength down 15%.
Honeywell was down 9% on the week after the Honeywell Aerospace investor day, underwhelmed with in-line top-line
growth, but muted 9% EBIT CAGR that could be lumpy between now and 2030. Airlines were also weak with higher oil
prices with AAL, UAL, ALK, as crude reversed course on the week.

Iran War and Oil Market Update - 45:24

Now, on the following page on the Iran update, um, world oil inventories are falling at a very rapid pace down to um, 6.8
um, billion barrels expected by September, uh, which would be an operational floor level. You know, there's minimum oil
required to keep uh pipes operating um at reasonable levels. Um, so here's a consequence of the Hormuz moves crisis
that actually matters and it printed last week. You know, China has been tapping deeper into its oil stock piles as its crude
imports hit a decade low. China is the largest oil importer on the earth. When it cannot get the barrels it wants um through
a disrupted straight, it does not simply pay up. It reaches into its strategic reserve and drains it. A strategic reserve is a
buffer. It exists precisely so a country can ride out a supply shock without the price exploding. China is now burning that
buffer. That is the quiet clock under this whole crisis. Every barrel China pulls from its reserves smooths the price today
and is gone tomorrow. The buffer is finite, the disruption is not. The market is watching the ceasefire headlines, it could
be, it should be watching the inventory. When the largest importer is draining its tanks and to keep the lights on, the V-
shape everyone is pricing in is not coming. So what happens to the price when the buffer runs out? Um, we'll likely see
China push for peace before then.

The US is investigating right now using for foreign, frozen Iran funds to assist other Middle East allies to rebuild after
Iranian strikes. And as a result, Iran is escalating the conflict also after Israel attacks Beirut and US breaks its ceasefire.
Besant today said that the Treasury will utilize all tools to allow Iranian assets to be made available to our Gulf allies to
support rebuilding and repairs for any future damage caused by Iran. Um, and uh, obviously they're very, they're very, um,
pissed about that, uh, because they expected, you know, 24 billion of capital to be, uh, released once they played, uh, you
know, towards peace with the US. Um, so you can read, you know, MB Ghalibaf's, um, discussion that, you know, there's
no ceasefire. A Hormuz letter talking about the IRGC now ready to execute Operation True Promise against Israel tonight.
So there's only there's an escalation going on on both sides.

Triggers for oil to go higher. The single most important tail risk is any new big disruption to the Strait of Hormuz and
damage to existing tankers, given 20% of global oil flows through there. The market is complacent about Middle East
escalation risk, which is being buffeted right now, uh, by lower Chinese imports. The US strategic petroleum reserve is also
no longer large enough to fully suppress a sustained geopolitical spike. The US Cushing inventories could get close to zero
this summer, which would result, which would result in a WTI spike, probably higher than Brent. We are weeks away from
problems getting much worse uh, into driving season and AC season. Salt caverns along the US Gulf Coast can't operate
below certain levels and Saudi spare capacity is not fully usable right now and Russian infrastructure has also been
damaged by Ukraine. Many macro funds were positioning for slowing growth and lower oil, while CTAs and systematic
traders were underweight energy, which has resulted in a big short squeeze.

Impact of AI, SpaceX, and Open AI on the Market - 48:51

On the following page, if you go to page 18, we discuss SpaceX and the market for a moment. If you remove AI, um, and
space basically, the SPX rallies uh, the SPX rally disappears since the start of the Iran War. Um, if you look at, uh, the
upcoming SpaceX IPO and OpenAI IPO, they will be the largest in history. Um, you'll see the SpaceX 1.8 trillion market cap
and the what's now an 84 billion offer size up from 75 billion. Um, now on OpenAI, the news that OpenAI is preparing to
file for an initial public offering very shortly, probably by September, suggests that the maker of ChatGPT aims to strike
while the iron is hot like SpaceX and Google, which is, you know, raising 84 billion last week as well. Um, the poster child
for AI plans to make a confidential filing with regulators as soon, uh, as this weekend and aims to go public by as early as
September. The Wall Street Journal reported, um, on the upcoming IPO with an anticipated surge in IPOs this year
including its fastest-growing rival Anthropic by December, there will be a scramble to make the most of investor appetite
for direct exposure to pure-play AI companies in public markets. Retail investors currently have to pick companies with AI
offerings in other parts of the supply chain, either in infrastructure such as semiconductor makers or cloud providers, or
on applications that sit on top of foundation models. Shares in the Mag 7 tech companies have quadrupled since the
OpenAI launch of ChatGPT in November of 2022. OpenAI was valued at 152 billion after completing its latest funding
round at the end of March. Reports have suggested that it could be planning to raise 60 billion in an IPO that would value it
more than 1 trillion. That would be the biggest IPO ever outside of SpaceX, twice as big as the Saudi Aramco IPO, which
valued, um, was valued at 25.6 billion, assuming no one else gets here first. Um, you can see that below in the diagram on
page 18.

Valuation expert Demodaran basically has been cautious, expressing caution on the SpaceX valuation. Um, you know, he
compares it to to Tesla. Um, he talks about how, um, you know, his DCF, um, gives a company value per share of like 97,
which is below the offering price of 130, even though in Hyperliquid it's now trading at 170, which you could see on page
20. Now, if I were JP Morgan's like private bank client who just got allocated a thousand shares of SPX at, um, 135, you
know, 135 grand, why wouldn't I immediately short an equivalent amount on Hyperliquid perps, uh, that were trading as
high as 200 and lock in the spread? The interesting part is the timing. The S-1 was filed on May 20th, so everyone had the
financials then. The 135 IPO anchor wasn't established until June 3rd, and the real negative price action in the perps
started after that, not after the S-1. To be clear, the perps still trades well above the IPO price, so this isn't a full
convergence, but the timing suggests the move lower was less about people suddenly reading the numbers and more
about the synthetic market finally getting a real-world anchor. You can see the hyper-dash, uh, print on page 20.

Memory Chip Sell-off Analysis - 52:18

Now, let's quickly talk about the memory sell-off, um, on Friday with MU, DRAM, Korean stocks taking a beating. So what
caused the sell-off in semi-space? Uh, the first catalyst was the SemiAnalysis report with Nvidia's Vera Rubin, uh, chip
using less system memory. So the SemiAnalysis article that shook the markets specifically about Nvidia's next-gen Vera
Rubin NVL72. What changed was the system-level DRAM capacity per rack is being configured at 28 terabytes of memory,
down from the previously expected 55. So almost down 50%. The second catalyst was that Korean's Korean market
because of high leverage, you know, the KOSPI crashed by about 6%. And then there was also a SemiAnalysis report on
Broadcom's soft guidance, uh, combined to create a violent sell-off in South Korea. The KOSPI plunged 5.54% to 8160,
which was the worst single-day drop in month months. SK Hynix, one of the world's biggest memory makers, was down
10%. Samsung Electronics was down 6%, and the Korean Won hit a 17-year low versus the US dollar, which compounded
the sell-off. The catalyst three was, um, Anthropic's 1A once a, uh, 1A AI, uh, builds itself report. This was published on
June 4th. So Anthropic basically dropped a bombshell report calling for the ability to slow down or pause frontier AI
development. Co-founder Jack Clark described AI development as having a powerful gas pedal but no brake pedal. They
proposed a coordinated global mechanism involving the US and China to provide the option to slower pause
development. The net effect is that these three catalysts are converging to create a question the AI trade narrative across
global capital markets, memory demand concerns, a leading AI lab calling for a slowdown, hot macro data keeping the Fed
hawkish, a potent potent cocktail for volatility, as you can see in the three graphs on page 21. The sell-off came after a
ferocious rally in memory stocks, especially in South Korea, and you can see the South Korean market being up 313% since
2025. Micron was down 13.5% on Friday, um, and you can see some summary about chip stocks below on page 22. Major
chip stocks sold off Friday to seal the sector's sharpest one-day loss in more than six years. The PHLX Semiconductor
Index (SOX) closed down 10.3% Friday afternoon, its largest one-day percentage decline since March 16, 2020, according to
Dow Jones.

Investors are considering the implication of a strong jobs report and weighing the potential impact of higher cost of capital
on future earnings growth and earnings misses like Broadcom. Investors are also cooling on momentum stocks and
considering the implications of a strong jobs report. Micron shares were obviously down 13%. Um, among notable losers,
you had MU down 13.3%, Intel down 11.3%, Advanced Micro Devices or AMD down 10.9%, Broadcom down 8%, Nvidia
down 6.2%, Marvell down 16.7% after Jensen Wang pumped the stock. "The fact that many non-tech sectors are holding
up reasonably well suggests that this is a semiconductor correction rather than a broad risk-off event," according to David
Nicholas, founder of Nicholas Wealth Management. Bernstein analyst Stacy Rasgon urged investors to have a little
perspective in light of the recent sell-off. "Nothing keeps going up every single day," he said. She said. Uh, the
semiconductor trade has become crowded as it is the only thing that's worked. Um, Bernstein added, and despite Friday's,
uh, deadline, the SOX index is still up 70% so far this year. In evaluating Friday's moves, I said that, um, investors looking
forward to high-profile initial public offerings might be taking some profits in their winning stocks so they can use that
money to buy shares of newly public traded companies when they become available to purchase. Brian Mulberry, who is
the chief market strategist at Zacks, said that the reasons behind Friday's chip sell-off were twofold. For one, a better than
expected jobs report was a key element for the Fed to raise rates, um, since the labor market looks durable enough to
absorb a higher cost of capital. A higher cost of capital would be a headwind to future growth. Broadcom's earnings
report earlier in the week also weighed on the broader chip sector. Investors were disappointed that Broadcom declined
to raise its forecast for more than 100 billion in AI chip revenue next year. It confused the market that growth expectations
didn't get a lift despite AI chip revenue jumping more than 143% from a year ago. Wedbush's analyst Matt Bryson told
MarketWatch that investors may increasingly be worried about shifting fundamentals in the memory market and the
prospect of a supply, uh, chain freeze coming to outstrip demand. As it stands, memory and storage makers have been
major beneficiaries of AI-driven supply shortages that have allowed them to raise prices. "There are two ways the supply-
demand dynamic could get screwed up," Bryson said. "Either you get demand destruction or you get more supply," which
will likely happen over time, but not this year.

Jensen Huang's Comments and Software Sector Rally - 57:28

Jensen Huang also kind of surprised the market and caused a big software rally early in the week where he basically said
that AI will create more uses for software. Um, you know, the following names were kind of attractive based on what he
said. Um, screenshot of his interview on page 24. So ServiceNow is basically working with Cadence and CrowdStrike and
Palantir, um, was named an enterprise AI partner. Um, Micron, you know, obviously should, despite the Vera Rubin sell-
off, should benefit from HBM4 memory demand. Um, it's the only US-listed HBM4 supplier for Vera Rubin. Uh, TSM, it's
everything starts with TSMC. The seven new chips that make up Vera Rubin take shape through hundreds of processing
steps in a 3-nanometer process where the sole manufacturer of all Vera Rubin chips is TSM. Uh, Dell obviously has done
very well, but on pullback, it could be interesting. Dell and CoreWeave, um, you know, are, uh, big supporters of the Vera
Rubin ML MVL72 engineering rack. Uh, CoreWeave, um, is a flagship AI cloud partner, um, that has been, you know, doing
a lot better. It has a path to cash flow positive. Microsoft, um, has also adopted the Vera Rubin ML MVL72 engineering
rack. Microsoft and Nvidia over the last three years have been also reinventing the PC. Cadence CDNS is the only
company given a full dedicated AI demo segment. CrowdStrike was named enterprise AI toolkit partner, so was Palantir,
so was SAP. Adobe uh has re-engineered the core of Photoshop and Premiere and saying they're going to release it for
RTX Spark. It's twice as fast and uses AI. Shopify was named as CoreWeave AI cloud as a AI cloud customer. Uh, it's trying
to reduce its costs, um, using AI. Google, um, you know, this is Nscale and their customers are British Telecom. It's
growing its um, cloud business. Its cloud backlog doubled. Um, and you know, there's been some negative sentiment
around the 80 billion equity raise, but overall, the company is growing rather rapidly. Um, ICE NYSE, um, Lynn Martin, the
president of the New York Stock Exchange has also partnered, uh, with the Vera CPU to deploy live NYSE real-time stream
processing. And then IBM, Red Hat, you know, obviously the Trump administration, you know, is was pumping IBM, um,
but Red Hat is adopting Nvidia's Open Shell enterprise runtime. Jensen had said last week that Marvell is the next $1
trillion company sending the shares up 25. percent after the sell-off on Friday, you know, the market quickly realized that
Jensen Huang had bought 2 billion of Marvell stock ahead of saying this.

Nvidia and X.AI Deal Structure Analysis - 1:00:37

Um, Nvidia and X.AI's deal structure funded by Athene, you know, Michael Burry's been very negative. Um, I think some
of his fears are a little bit overblown, but you can see how the SPV is constructed on page 26. Effectively, Nvidia, the
chipmaker sells 100k GB200 GPUs worth 5 billion to the SPV and books it as a sale. The Valor SPV is a shell entity that
holds legal title to 100 plus K GPUs with no operating business and then gives its equity to Nvidia. Then there's a 1.9
billion injection by the sponsor from Nvidia and its affiliates into the SPV. The 5.4 billion of GPUs serve as collateral, uh, for
Apollo, um, which, you know, holds, uh, you know, which lends money, um, to the SPV. Apollo then earns fees for the
arrangement and structuring of the deal, asset management, sponsoring, and servicing. Then, you know, there's a debt
securitization of the, uh, debt used to fund the SPV and the debt is packaged into securities and then sold to investors with
Apollo reducing its risk. There's 3.5 billion of debt securities then sold, um, to Athene, which is Apollo's insurer, which
buys the debt securities and holds them in its portfolio. Um, Burry is basically saying that there are some alarms here
because Athene has 74 billion of total reserves, uh, but 217 billion has been shifted into a Bermuda captive outside the
US, which is a source of a lot of leverage. And 103 billion or 34.7% of the company's assets are in level three assets, which
are basically private credit with no observable market price. After Athene, um, buys the debt securities, it then sells, um,
annuities to its holders. So American retirees essentially bear the risk of these, um, Nvidia GPU SPVs. And so effectively,
you know, the net effect is Nvidia books 5.4 billion as a completed sale. Apollo collects fees across origination,
structuring, distribution and servicing from its retirees. X.AI gets 5.4 billion of computing power with no asset or debt on
its balance sheet. And then retirees sit at the bottom of a 16 times leverage structure holding 103 billion of unpriced
assets, carrying the risk unknowingly. So the transaction is real and the figures check out. Apollo led a 3.5 billion capital
solution for Valor compute infrastructure to fund a 5.4 billion purchase of GB 200 GPUs, leased to X.AI on a triple net
lease structure. I know it's complicated. Nvidia went in as an anchor LP, all publicly disclosed, but the accounting isn't the
isn't erroneous, um, you know, fully. Burry's thread oversells two things. Number one, Nvidia's revenue, selling to an SPV
is fine as long as it's a real sale. The question under ASC 606, um, accounting, which is the US revenue standard, is whether
control is actually transferred. If VCI bears the risk and rewards, then Nvidia books the sale legitimately. The real issue is
the 1.9 billion Nvidia then ploughs back into VCI as an investor. That is the round trip. So net, Nvidia took in roughly 3.5
billion of outside cash, but booked 5.4 billion in revenue. So there is some accounting funkiness here, but it isn't fully
funky. If part of your sale is funded by capital you reinjected, that portion isn't a sale. The honest treatment is either the
net of the 1.9 billion off the sale transaction price, so 3.5 billion net, or you run a variable interest entity analysis and
consolidate VCI into into Nvidia. Recognizing gross revenue on a round trip capital is the potential weak spot of this
analysis. On the legally invisible, um, this is rhetoric. The chip sit on VCI's balance sheet, X.AI carries an ROU asset and a
lease liability under ASC 842, which is the US leasing accounting standard. Nothing actually vanishes. It's held by an entity
just nobody consolidates, and whether that non-consolidation is correct is the VIE question above. On level three assets,
um, no outside party can verify what they're worth. That's also wrong. Level three means no observable inputs for that
specific asset like private credit. It's not unverifiable, it's just very hard to decipher. We typically always brought in
valuation specialists, particularly for high-risk material and what you do to value private credit and other level three assets
is you use observable comps and secondary, you know, um, GPU prices in the case of valuing GPUs as model inputs. And
auditors treat it as a critical audit matter. Uh, it gets more scrutiny, uh, not less. So it's not like the GPUs can't be valued,
um, and the private credit on Apollo's balance sheet can't be valued. It can, you can estimate it. The legitimate concern is
smaller than this post lets on. So level three marks are management estimates exposed to optimistic bias, 34.7%
concentration is high for a retail annuity backing and that sits on top of 16.6 times leverage in a Bermuda captive outside
the US statutory oversight. The stack GPU residual value risk on a multi-year lease and that's the main concern. So Burry's
substance is partly defensible. The retirees are unknowingly carrying invisible risk. Packaging is sensationalized. Policy
holders hold fixed contractual claims and their exposure to is to Athene's insolvency, not directly to GPU residuals.
However, if Athene has compounded issues due to software, private credit, etc, it could be that retirees are exposed. Uh,
so in summary, auditors need to test whether the sale is overstated by the 1.9 billion round trip and apply extra scrutiny to
the unobservable level three inputs. I'd hate to be the audit partner signing on these off these transactions, particularly
given the public interest and frequency of similar transactions. You can see the Valor X.AI compute transaction
summarized on page 27.

CPU Shortage Discussion - 1:06:47

And then we'll talk about very briefly the CPU shortage, but before that, I just need to get some more water.

So, um, if you look at the CPU shortage, um, this is this discussion is actually, um, quite interesting. Um, just go back into
into this. Um, so agents make GPUs need more CPUs, and CPUs need scarce memory, wafers, and packaging. So the trade
is effectively, um, bottlenecked stack around the CPU. The CPU trade is actually an extension of the memory trade. Um, at
the end of March this year, a CPU company launched a new CPU. It was called the AGI CPU.
So, okay, a CPU company launching a new CPU, what's the big deal? Um, except that this one came from Arm, which we all
know, ARM, a company that in 35 years had never really sold CPUs. Arm is usually the third name in the data center CPU
conversation alongside Intel and AMD. But unlike the other two, Arm doesn't actually build chips. It designs CPU
architectures and license them to companies that do the building like AWS Gravitron, Microsoft Cobalt, Google Axeon,
Nvidia Grace, and so on. So when a 35-year-old licensing business pivots to selling finished silicon directly into the data
center, it's a signal that shows you how much demand there actually is for CPUs.

So when Arm announced the launch on March 24th, they said that it already had 1 billion of customer demand lined up for
the new product. And six weeks later, they said that that jumped to 2 billion. Two X in six weeks. Does that mean the new
CPU is twice as good as everyone thought? Not really. It's more that likely that data center CPU demand is just so high
right now that the customers are grabbing anything they can get. And Arm is the only company that suddenly looks
different in the CPU business. It's also why Intel rallied.

Days before the earnings call last week, Nvidia had delivered the first standalone Vera CPUs to its partners like SpaceX,
OpenAI, and Anthropic and others making this chip's move into production. When two companies that had no business
selling data center CPUs start doing exactly that in the same quarter, the market is telling you something. CPU has been
critical has become critical to modern AI infrastructure in a way that simply wasn't there a few years ago. And the thing
driving it is is the explosive growth of agentic AI.

Industry-Wide CPU Demand Signal - 1:11:19

All four major CPU names, um, revised the opportunity upward. All four pointed to agentic workloads and the CPU to GPU
ratio language moved in the same direction from roughly one to eight historically towards parity. So the demand signal is
quite clear. The supply side was already flashing. So that's the setup. AI agents turn the CPU from a background host into a
scaling layer for the data center. That is driving an explosion in server CPU demand. The problem is that this demand runs
into a supply chain already stretched across leading edge wafers, DRAM, and advanced packaging. Meaningful relief
probably doesn't arrive before 2028, believe it or not.

So the clean trade is simply simply not to buy CPU vendors. It's to own the scarce inputs. TSM for manufacturing, Micron,
Samsung, SK Hynix. Um, which we'll discuss later. And packaging players with a clear linkage to CPU demand such as
Amkor, ASC, which we thought had become overvalued before.

So on page 29, you can read why agentic AI is a CPU story. An agentic workload looks structurally different today. An agent
reasons across a long task, it calls for tools like search, code execution, browser, automation, database queries, pulls in
and updates states across many steps, makes branching decisions based on intermediate results, and coordinates across
sub-agents and processes. So you know, in the following few pages, I just have still have a lot to go through the through
the call, but you guys can read about Arm's AGI CPU, um with 136 cores and why demand has been rising.

Supply Chain Bottlenecks and Investment Thesis - 1:13:24

You can read about how AI racks are getting CPU heavy with an interesting diagram on page 35. You can read about the
demand signal. Um looking at the market itself itself bottom up sizings put agentic AI at roughly 40 billion in 2026 and 140
billion by 2030. You can look at Claude code demand going vertical, the graph on page 32. You can look at who actually
builds the CPUs. Um and on Intel on page 33. We talk about fabless designers like AMD. Um and the memory squeeze on
page 34, the HBM premium versus a DRAM collapse, um on page 35.

You also talk about wafers, the allocation trade-off on page 35. You know, the relative value between a GPU and a CPU on
page 36. Packaging bottlenecks on page 36. And then the conclusion that the GPU is still the center of the AI data center
and it remains the biggest power draw, the biggest capex line, and the hardest component to source at scale. Agentic AI
changes what has to sit around the GPU. Agents call tools, they search files, they write code, they run tests, they
coordinate sub-agents, they hold state across long tasks. That turns the CPU from a host processor into a scaling layer.

Once that happens, the supply chain gets harder. More agents mean more CPUs, more CPUs mean more DRAM, more
CPUs mean more leading-edge wafers, and more CPUs also mean more packaging substrates, IO and power delivery. The
next bottleneck is unlikely to be one clean component. It will likely be the collision of every layer of the rack trying to scale
at once. This is why the cleanest trade is not only buy CPU vendors, the cleaner trade is to own the scarce inputs. And we
still think that TSM is one of the best trades, uh, in AI.

Market Commentary: Morgan Stanley, Nvidia, and Google - 1:15:17

Morgan Stanley's takeaways from meetings in Taiwan for semiconductors. Um it points to Nvidia data being still very
strong on Ruben demand, Vera Ruben demand. AMD and Intel continue to see strong service CPU trends. Um so, you
know, Blackwell production is also re-accelerating. The $20 billion Nvidia CPU target also seems reasonable. So our CPU
discussion earlier explains why Morgan Stanley is so bullish bullish on CPUs. Um and that AMD continues to be a
potential share gain gainer in the CPU with the x86. That's on page 39.

Then on page 40, they talk about memory shortages will still persist for some time, which makes me think that the selloff
last week, you know, was because of overvaluation, it doesn't necessarily mean that the memory trade is over. Um, data
center construction spending has eclipsed 50 billion more than public sector infrastructure spend. You know, Google's
equity raise to fund AI infrastructure is on page 42. You know, Google was by far the most asked about name last week
with a lot of confusion around the equity raise. From Morgan Stanley's conversations, their interpretation boils down to
either management believing AI capex will end up being meaningfully higher than what the market is currently modeling.
Remember, guidance was only qualitative for 2027 capex, but conversations suggest buy-side expectations are 270 to 280
billion for Google AI capex, which which raises questions around the ultimate ROIC.

Regardless, the knee-jerk reaction here is negative and we all get why. Though a few other speculations around the move
being tossed around are worth flagging. Maybe management is trying to get ahead of the onslaught of IPOs in the second
half of the year with OpenAI, Anthropic, and SpaceX proactively pulling liquidity out of the market and using the
proceeds to further secure more ever so stretched uh, you know, supply chain memory and GPU assets. That's not a wild
thought. Maybe they're anticipating another step function reaction in demand for GCP or Gemini.

Um, and we're, you know, the market's a little bit worried because we're already starting to see some early cracks in token
maxing with Uber, Amazon, and Duolingo all talking about trying to be more efficient with their AI spend.

Meta's AI Capacity and Strategy - 1:17:34

On uh page 44, we talk about Meta's gigawatt capacity coming online. They have an unbelievable amount of AI capacity
coming online. Um, they were also hacked when it comes to Instagram and we saw some insider selling which caused the
stock to sell off. Um, Meta's total effective 4 gigawatts added in 2026 and '27, including hyperscale deals, speaks to the
importance of product innovation, incremental revenue. This gives them the capacity to actually focus on developing their
own AI. Right now their product is basically shit. But, you know, they do have the capacity to improve it and we'll see how
that plays out. Um, Meta's latest push on global subscription tiers is is talked about on page 45.

HP Earnings and the AI Labor Debate - 1:18:14

HP earnings blew it out of the water, which we should have anticipated after Dell last week, but HPE had already rallied so
much. We described those earnings on page 46. The AI labor debate, um, is on page 47, which talks about how OpenAI
CEO Sam Altman admits he's wrong. Yann LeCun, uh professor at NYU and now head of AMI Labs, you know, he said we
we told you not to listen to Sam and Dario about AI destroying jobs right away. Um and basically, you know, consumer uh
views of AI companies were getting so negative that uh both Dario and Sam basically had to cave and said, hey, this is not
true, it'll take longer, you know, it'll make people more efficient. Um it's just a a temporarily change in the narrative. Um,
so you can read about the two reversals in vindication on Sam's view and what the data says uh about tech layoffs on
page 48.

Um, and there's a really interesting case study about how Pizza Hut, the largest Pizza Hut franchisee is suing the
company for, you know, AI effectively making delivery slower on page 49.

Analysis of the Pfizer Investment Case - 1:19:26
And I just wanted to spend a few minutes to quickly talk about um the Google uh equity raise, which you could read about
actually on page 51 and uh US savings being at the lowest at a multidecade low of low of 2.6%, which was you know, kind
of showing bubble territory and and excessive speculation. The Pfizer model, um, is on page 53 and 54. And I just wanted
to share that, um, you should go through through the model. Um I am going to, you know, work on more of a write-up on
this. Um but effectively, we think the stock's worth, you know, between $30 and $32 a share. It pays a 6.6% dividend. Uh
it's relatively defensive. You know, we laid out its um, its backlog, um, of new drugs coming out. Um, effectively modeled
out its weighted average cost of capital and did a perpetuity sensitivity. This stock is basically implying no growth, um, at
$26 a share. Um and what and what we put in was we put in a 2% terminal growth and a 7% WACC and we think that, you
know, if the company shows any signs of growing, you know, it's worth well above $30 a share. We also did a probability
adjusted analysis on with three scenarios. You know, a optimistic case, um, where we think the stock would be worth $41
a share, a base case um, you know, at 28, and then a bear case where if the the company's not able to to prove um,
growth to the market, it falls to about $20 a share. So the up-down is pretty compelling. Um, it's also, it also rallied on
Friday. There is demand for financials and healthcare, um, in the SPX and industrials, uh as tech sells off. Um so we do
think that this is a relatively low beta, has like a 0.7 beta to the market name to add to the portfolio if you're concerned
about having too much exposure to to tech. You know, even the SPX has about 20% exposure to semiconductors and 40%
exposure to tech. So we're all probably overexposed here.

Q&A Section - 1:21:22

Um, let me go just quickly answer some questions. I do have an appointment to head to in a bit.

Appropriate Sizing for SpaceX IPO - 1:21:27

I would say an appropriate basis point sizing for SpaceX IPO shares would be, you know, basically, um, 20 bips. You know,
our price target for SATs, um, you know, is is over $150 a share after the resizing of the IPO. And, you know, if you look at
Hyperdash, SpaceX is trading well above um, 135. So you could see a small pop. I just wouldn't get too big because um, I
do think that, you know, the valuation is very excessive at like 200 and 60 billion uh 260 times forward EBITDA. Um, you
know, I do think there will be a retail rush towards it as as, you know, the IPO um, requirements have fallen. Um, but um,
you know, I just wouldn't want to be too big, too big in something like this because then you have the OpenAI IPO and
then the Anthropic IPO. I just don't know how much demand there is for these really big IPOs. I think Anthropic is the
most interesting. Um but on SATs, you know, just where Hyperdash is right now, if you allocate to the IPO, you might see a
pop. Um that will be, you know, likely sold off into the IPO unlock uh six months later. So I hope you guys um enjoyed the
call. Um the recording will be up very shortly and uh we'll have some other updates during the week.

Date of Transcript: 2024-05-24
Filename: ssn_transcript_v6 - 2026-01-26 - @DA.mp3
Model (may not be accurate): gemini-1.5-pro-001
Prompt: ssn_transcript_v6 - 2026-01-26 - @DA
QA Check: Completed