Jay Singh — Weekly SSR: the Mag7 capex-burn selloff, Micron's take-or-pay reset, Sea Limited cheap, Rithm Series F prefs at 9%, BTBT at a 43% NAV discount, the Alamos (AGI) follow-up

Last week a small earthquake hit one of Alamos's four gold mines (Young-Davidson), knocking out part of one tunnel. The stock fell by a couple of billion dollars — but the actual repair is tiny (~$25 million, a few months of clearing rock; nobody hurt, no machinery broken), and that mine is only about 15% of the company's worth. The crown-jewel mine, Island Gold, is 60% of the value and was completely fine — its ore is five times richer and far cheaper to dig.

After calling the analysts who cover it (RBC, Jefferies, Evercore), Singh values the whole company mine-by-mine at about $48 a share. It used to trade at a premium to rival miners; now it trades at a discount to them — even cheaper than riskier emerging-market miners. He's treating it as a long-term investment, not a quick trade: it needs two or three clean quarters to recover trust, and a planned doubling of its processing mill should drive ~30%-a-year revenue growth.

Sea Limited is a Southeast-Asian internet conglomerate: a video-game business (Garena), the region's dominant online shopping platform (Shopee), and a digital-payments arm (Sea Money / Monee). Singh says it has "never been as cheap" as today after a selloff driven by fears of TikTok shopping and AI eating into its business.

Adding up the three parts separately and subtracting a discount for being a holding company, he gets roughly $187 a share versus about $91 today — around 105% upside — and potentially much more if Shopee can raise prices the way Amazon did over time. A new free AI assistant, "Migoo," is meant to keep users inside Sea's apps. It has ~$10 billion of cash (over a fifth of its value), no debt, and is still growing ~19% a year.

A "preferred" share is a hybrid that pays a fixed dividend and sits ahead of the regular stock if anything goes wrong. Rithm Capital is a diversified, profitable financial firm (mortgage servicing, which actually does better when rates rise, plus a property arm). Its Series F preferreds yield about 9.2% because they trade a bit below their $25 face value, and they pay every quarter.

Singh's "common-sense" point: the ordinary stock yields 10.6%, so for just 1.4% less you can own the preferred and have the entire $5.2 billion of common stock acting as a cushion beneath you before your dividend is ever at risk. He rates it a low-risk income idea (2 out of 5) and is adding it this week. It's available to any retail investor.

Rithm is the company that issues the preferred above — a roughly $5 billion diversified asset manager trading at about five times earnings with the common paying ~10.6%. Singh likes the business (its mortgage-servicing book is a natural hedge against rising rates, and it has little exposure to the troubled private-credit market) and thinks it's safer than a pure mortgage REIT like Redwood — but he prefers owning the safer preferred to the common here.

Bit Digital used to mine Bitcoin; it shifted into cash, Ethereum, and AI infrastructure, and it owns 70% of a company called White Fiber. The quirk: White Fiber alone is worth $1.5 billion on the market, yet the entire Bit Digital is valued at only ~$790 million — so you're effectively buying its stake in White Fiber (plus a pile of cash and crypto) for a ~43% discount. Singh added a small position betting that gap closes as White Fiber grows.

White Fiber rents out high-powered computing for AI — GPU hosting and data-center colocation (similar to DGXX) — and importantly it carries no debt. It's the valuable asset hidden inside Bit Digital. If White Fiber delivers on its profit targets, the discount on its parent (BTBT) should disappear.

Micron makes memory chips. For 47 years that was a brutal boom-bust business: prices rise, Micron makes money; prices fall, it loses money. This quarter it earned more profit than Nvidia did a year ago (85% gross margins; ~$50 billion of revenue guided for next quarter).

The real news is 16 "take-or-pay" contracts — five-year deals (2026-2030) where customers must buy set volumes or pay penalties, with a guaranteed minimum price set so high that even a bad memory downturn would leave Micron's margins above any past peak. Customers even pre-paid $22 billion to lock in supply. That converts the most volatile business in chips into a steady, high-margin one. (The bear case: Chinese funds argue memory has "no moat" and rivals can build capacity over time.)

While reviewing the gold-miner comparison sheet for Alamos, Singh flagged Hudbay — a copper miner — as "quite interesting" and a name he'd be buying for long-term exposure to industrial copper demand.

Redwood is a mortgage company that buys and packages jumbo home loans. It fell to about $4.81 on rate swings and a weak quarter, but that's roughly a 33% discount to its $7.12 book value, and its earnings still cover the dividend — a ~15% yield. Singh owns it, sold higher, and has been buying it back; he just needs interest rates to fall for the value to be realized.

MicroStrategy (now "Strategy") issues separately-traded preferred shares to fund its Bitcoin buying. When crypto fell, these dropped sharply — STRC by about 25 points. Singh started buying at an average price in the mid-to-high 70s (face value 100), so simply returning to par plus the dividend is over a 30% return. His bet: Michael Saylor can still sell common stock to keep paying these dividends. He sizes it small — it's not without risk.

A cleaner version of the MicroStrategy preferred trade: this issuer has no debt and only preferred shares, so there's less ahead of you in line. The prefs yield about 15% and were buyable in the low 80s. Singh likes it but warns he "wouldn't be too aggressive," because renewed war fears could trigger another crypto selloff.

Microsoft has fallen about 35% from its high — its worst June since 2008 — caught in the broad selloff of big tech that's pouring cash into AI. It actually spends less on data centers than its peers, and famed bear Michael Burry just bought long-dated call options on it. Singh would add gradually into this weakness, while watching two overhangs: its biggest cloud customer (OpenAI) is delaying its IPO, and memory costs are forcing Xbox price hikes.

Of the big AI spenders, Meta is the one Singh is most wary of. It's down about 33% and is spending more than its entire cash flow on AI, with a weak free-cash-flow outlook. If it tried to raise $80 billion in stock the way Google did, he thinks "the market would really hammer it" — and the shares already dropped just on rumors of a raise.

Oracle is "the worst off" of the group: it has the weakest cash flow, its debt-insurance cost (credit-default swaps) is rising again, and it's cutting staff — headcount down to ~141,000 from ~160,000 a year ago, with more layoffs coming — all to fund its AI data-center push. A name on the wrong side of the capex-burn trade.

SpaceX has become a new "hyperscaler," renting out data centers (to Anthropic and Google) and raising $20 billion of debt. Singh isn't shorting the stock — he's eyeing its longest-dated bonds (maturing 2056) as a cheap hedge against the whole AI boom: because very long bonds are extremely sensitive to interest-rate moves, even a small rise in its borrowing cost could knock ~30% off those bonds' price.

Not a buy yet, but a name to watch. Domino's has crashed from ~$500 to the high-$200s on fears that weight-loss drugs (GLP-1s) will curb eating out and on slowing same-store sales — even though it keeps taking market share and has raised its dividend nearly every year since 2016. Singh calls it the best delivery-pizza operator in the US and says "there will be a time to add" as it approaches deep-value levels.

An unusual windfall: a hedge fund (Pentwater) had to hand Avis a proposed ~$650 million in cash under the "short-swing profit" rule — a law that forces big insiders who buy and sell a stock within six months to give back the gains. The fund sold too soon after a massive squeeze, so it owes the profits to the company. Avis rose on the news; Singh had "interesting option trades" around it.

Varonis is a cybersecurity company that is "effectively selling itself" — it confirmed it's exploring a sale after private-equity firms (Blackstone, Thoma Bravo, Vista) showed interest, and the stock jumped ~10%. Singh is tracking it as a takeover/arbitrage situation.

Singh prefers S&P Global and Moody's — the dominant, wide-moat ratings-and-data duopoly — over FactSet, which he doesn't like. There's also a catalyst: S&P is spinning off its automotive-data arm (Mobility Global / MGGL, anchored by CARFAX), which just began trading on a "when-issued" basis.