In short: Sechan's final trade (42:05). "Up relatively little year to date, it's re-rated a bit lately, 60% revenue growth this year, 60% next. I think it's a relatively attractive valuation at 20 times."
Broadcom designs custom AI chips and networking chips for the big cloud companies. Sechan's point is about price versus growth: revenue is growing around 60% a year, yet the stock trades at about 20 times earnings and has lagged this year, so he thinks investors are paying a reasonable price for a very fast-growing business.
In short: Named by Ethridge (9:02) alongside AMD and NVIDIA as the chip names that led before the rotation.
In short: Weiss's exhibit A for not putting new money into semis — a falling knife after two misses. "I just don't still don't believe you put new money in here… Let me just give this one, Broadcom. Broadcom missed, right? So they missed arguably this time, they missed last time… they reported good numbers, guidance, squishy things in their margins. The stock has never bounced. It bounced for a brief time… I bought it after the prior quarter, bounced a little bit, sold it flat and it traded at 410. Now it's back down to… below where it was before, 346. To me, that's the message in the market. You can pick a falling knife here, particularly when you still have valuations that are too high." Santoli: somebody wants to own the hyperscalers "rather than Broadcom."
Broadcom designs custom AI chips and networking hardware for the big cloud companies. Weiss's complaint is not that the business is bad but that the stock has stopped responding to good news: it disappointed on two straight reports, and even when numbers looked fine, investors picked at guidance and margins.
He bought after one report, it barely bounced, he sold at breakeven near $410, and it is now around $346. To him that is the market telling you something — trying to buy a stock that keeps falling ("catching a falling knife") is dangerous when the valuation is still high.
In short: Subscriber view, passing: hyperscalers "hate this dynamic" of NVIDIA dependence and "are responding by developing their own silicon, working with Broadcom, Marvell, and others, but this takes time."
13:00Hence, the strategic logic of acquiring Hugging Face, which Nvidia acquired, I believe, last week. Hyperscalers hate this dynamic. They are responding by developing their own silicon, working with Broadcom, Marvell, and others, but this takes time. His point here is that the hyperscalers know that they are overly dependent on Nvidia, and they hate it, and they are looking for alternatives. I continue.
In short: Bought on the print, with the arithmetic stated. "We bought a little bit of Broadcom. Hock Tan's actually guiding to $30 of EPS by fiscal 28. On a 20 multiple, that implies AVGO could almost double. The company's growing at 50%. Now imagine they do 40 or 50 of EPS by 2028, 2029. This stock is basically pricing in that growth is peaking." The quarter: revenue ~$30B vs $29.25B consensus, EPS $3.32 vs $3.21, free cash flow $13.7B at a 46% FCF margin. The guide is the story: "the hyperscaler lock in — Broadcom secured commitments to enable over 20 gigawatts of compute capacity over its top customer base including Google, Anthropic, Meta and OpenAI through 2028," with FY26 AI revenue guided at $58B, FY27 doubling to $115B, and doubling again to $230B by 2028 — "which is quite aggressive when it comes to guidance." Q4 guidance of $34.8B implies a 93% sequential step up with Q4 AI revenue of $21.7B, +236% y/y. His reaction to the tape: "so we bought some Broadcom. I don't know why the market was not happier." Later, on the supply-constraint theme: "Nvidia is guiding to 70% growth. AVGO is guiding to 100% growth. Both are supply constrained and saying real demand is higher, but the market is not giving them credit." (Morgan Stanley's caveat, quoted: the $115B FY27 number came "slightly below the firm's $120 billion estimate," still in line with management's prior "well above $100 billion.")
Broadcom designs custom chips for the handful of companies building the world's artificial-intelligence data centres. Unlike Nvidia, which sells the same general-purpose graphics processor to everyone, Broadcom co-designs a bespoke chip for a single customer — which makes each relationship stickier and longer-dated.
The results were strong: about $30 billion of revenue against expectations of $29.25 billion, earnings of $3.32 a share against $3.21, and $13.7 billion of free cash flow, meaning 46 cents of every revenue dollar came out the other end as spendable cash. But the forecast is what matters. Management says its AI revenue will be $58 billion this financial year, roughly double to $115 billion next year, and double again to $230 billion by 2028 — underpinned by commitments to supply chips for more than 20 gigawatts of computing capacity for Google, Anthropic, Meta and OpenAI.
Singh's sum is simple. The chief executive is guiding to $30 of earnings per share by fiscal 2028. Put a fairly ordinary multiple of 20 on that and the stock roughly doubles — and that assumes growth slows, when the company is currently compounding at 50% a year. "This stock is basically pricing in that growth is peaking." He bought some, and was puzzled by the reaction: "I don't know why the market was not happier."
Full passage: premium transcript (PDF).
In short: A short-term reaction call on a print that had just come out, offered as evidence for the bears: "your viewers will probably see what's happened with Broadcom, which reported after the close, but the numbers were nowhere near what people thought they would be. And so that stock's going to go down probably tomorrow, unless something is changing on the conference call."
Broadcom designs networking chips and custom AI chips for the big tech companies. Its earnings came out while this interview was being recorded, and Niles' read was blunt: the numbers "were nowhere near what people thought they would be," so the stock was likely to fall the next day unless management said something reassuring on the call.
This is a short-term reaction, not a long-term verdict on the company. He uses it as evidence for the bear side of the argument — if AI chip suppliers start missing expectations, it suggests data-center spending is pausing, which fits his view that the next couple of months are risky and cash is the safer place to wait.
38:13And they're pointing at things like you brought up earlier which is hey token costs have gone down 50% from the end of May. If they stay on that pace and token usage doesn't go up, we could have a real problem. And the data centers, if you can't put the chips somewhere, you could have, I mean your viewers will probably see what's happened with Broadcom, which reported after the close, but the numbers were nowhere near what people thought they would be.
In short: Appears purely as a concentration exhibit, not a call: "I have a student who showed me his portfolio. It's 50% Nvidia and 50% Broadcom and that was his entire portfolio. And my suspicion is that's the case with a lot of retail investors in the US." The point is the holder, not the company — "they were the darling stocks for a long time. Everybody piled into them."
Broadcom makes networking and custom chips, and designs the bespoke AI accelerators that large cloud companies use as an alternative to buying everything from Nvidia.
Here it is not a company at all — it is the other half of a story about crowding. A college student showed Dillian a portfolio that was 50% Nvidia and 50% Broadcom, with nothing else in it, and Dillian's read is that this is roughly what a lot of American retail investors now own.
He is careful not to sneer at it: "I hesitate to use the word dumb money because they've been right." The concern is behavioural rather than analytical — concentrated holders who bought a long move are the least likely group to sell near the top, which is what turns an ordinary drawdown into a disorderly one.
10:07I teach college students, right? And I have a student who showed me his portfolio. It's 50% Nvidia and 50% Broadcom and that was his entire portfolio. And my suspicion is that's the case with a lot of retail investors in the US. They were the darling stocks for a long time.
In short: The episode's designated test of the momentum factor — and Terranova's answer is that there is no trade. The damage first: the stock is down 5% on the month and "in the SMH's top 10 holdings, Broadcom's the worst performer… down 23½% over the last three months," with a story out today saying it needs "an NVIDIA-like earnings report to stop what has been a $520 billion skid." Terranova reconstructs the break: "the last earnings report on June 4th signaled an inflection point… they missed by $1 billion in their AI revenue — 16 billion instead of 17 billion. The stock goes down 12% the next day." On valuation he gives both anchors: "it's 21 times forward now. It was 32 times forward in June. The 10 year average is 18." And then the refusal: "sometimes there's no trade, and I just don't see the setup in this for there to be a specific trade related to Broadcom… but if they come in tonight and deliver 200% earnings growth and 85% revenue growth and that's not enough, then that AI momentum trade is going to continue to sit silent." Lebenthal names what would change it: Broadcom has to say the pie is big enough that competition doesn't matter.
Broadcom designs custom AI chips for the big cloud companies (notably Alphabet) alongside its networking business. It reports after the close, and Wapner sets it up as a live read on whether the momentum trade is still alive.
Terranova's account of what broke is precise and dated: on June 4th the company missed its AI revenue number by $1 billion — $16 billion instead of $17 billion — and the stock fell 12% the next day. The reason a small percentage miss did that much damage is that the number was the evidence for the whole story; when the evidence wobbles, the multiple people were willing to pay collapses. It has: 21 times forward earnings now, against 32 times in June, and a ten-year average of 18. Cheaper, but not cheap.
His conclusion is the part most worth copying, because it is the option almost nobody exercises on television: "sometimes there's no trade." He is not bullish and not bearish; he is saying the setup does not offer him an edge, so he will not manufacture one. And he attaches a falsifiable test to it — if 200% earnings growth and 85% revenue growth is not enough to move the stock, then the AI momentum trade is genuinely dead for now, and that tells you something about every other name in the group.
In short: Reports "tomorrow" as they tape — "less of a bellwether than Nvidia," but one of the trio (with NVDA and MU) still delivering the capex-spend numbers the market is looking for. One of the two names she added to the Mag 7 to form her "Neural 9."
Broadcom designs custom AI chips and networking silicon for the big cloud companies, and it reports the day after this interview was taped. She calls it "less of a bellwether than Nvidia" — useful, but not the read on the whole AI complex that Nvidia's number is.
It matters here mainly because it is one of the two names she added to the "Mag 7" to build the group she tracks daily, the Neural 9. Along with Nvidia and Micron it keeps producing the capital-spending numbers that justify the AI trade — which is precisely why she is watching for the first crack.
27:59bring forth numbers that we're looking for, the capex spend numbers that we're looking for. It's hard to extrapolate that with an infinity sign. And at some point there's going to be some sort of miss. And I think what's interesting these days in this environment of just unbelievably strong earnings growth is number one, when do we actually hit the inflection point? Because I'm fond of saying and have for my 40 years doing this, better or worse often matters more than good or bad.
In short: Reports Wednesday, reiterated sector perform at RBC with a $400 target, and Wapner floats it as the week's next catalyst "after NVIDIA and after Marvell." Harrington shuts it down: "if NVIDIA couldn't be a catalyst, why would Broadcom be?… I also don't think it's as broad reaching. We were so reliant on NVIDIA to say exactly what they did and let that trickle through — and that was there." Wapner's counter is that the guidance bar is less well telegraphed: "how much does Broadcom raise its guidance? We felt like we knew we'd already checked all the NVIDIA boxes before they even reported. I don't think this is as cut and dry." A committee member owns it.
Broadcom reports Wednesday and is the week's next big AI test — it designs the custom chips large cloud companies use as an alternative to Nvidia's. RBC repeated a neutral rating today with a $400 target.
Harrington's dismissal is the memorable line: if Nvidia's spectacular numbers could not move the market, why would Broadcom's? Wapner's counter is worth holding onto, though: everyone had Nvidia's quarter pre-modelled, whereas how much Broadcom lifts its own forecast is genuinely less well telegraphed — so there is more room for surprise, in either direction.
In short: The clean read-through from Nvidia's guide, stated twice. "We have Broadcom next week. Based on Nvidia's outlook, you would think that Broadcom would beat and raise its outlook as well," and again after the print: "this should be also good for Broadcom this week." The custom-silicon evidence is his own: OpenAI's internal testing showed its Broadcom-built Jalapeño (elsewhere "Palomino") inference chip outperformed Nvidia's GB300 in throughput per watt and latency at 700 watts, with deployment "targeted for later this year" — the caveat being it was benchmarked against GB300, not Vera Rubin. "We're optimistic for Broadcom results this week, but we'll see." The offsetting fact he keeps in view: Broadcom is one of the seven names inside the $3.1 trillion of hyperscaler off-balance-sheet arrangements, and Goldman's Delta One desk names "NVDA/Broadcom CDS has widened materially as the market focuses on AI financing backstops and off-balance-sheet commitments."
Broadcom designs custom chips for individual customers rather than selling one general-purpose product to everyone. Its biggest customer project is OpenAI's own inference chip — the processor that runs a trained AI model to answer queries, as opposed to the far more expensive process of training it in the first place.
This week OpenAI's internal testing showed that chip beating Nvidia's GB300 on two measures that decide data-centre economics: work done per watt of electricity, and how quickly it responds. At 700 watts, with deployment planned for later this year. The caveat is that it was compared against GB300 rather than Nvidia's newer Vera Rubin generation.
Singh's near-term reasoning is simpler and mechanical: Broadcom reports next week, and Nvidia has just guided to 70% revenue growth. "Based on Nvidia's outlook, you would think that Broadcom would beat and raise its outlook as well." Whatever demand is filling Nvidia's order book is filling Broadcom's too.
The offsetting fact he keeps beside it is a credit one. Broadcom is among the seven companies carrying $3.1 trillion of commitments that sit outside their published balance sheets, and Goldman's trading desk notes that Nvidia's and Broadcom's default-insurance costs have widened sharply as the market focuses on how all this construction is being financed. The equity is optimistic; the credit market is asking questions.
Full passage: premium transcript (PDF).
In short: Peak-date evidence, not a stance: "even Broadcom and Micron peaked different weeks a few months ago" — the point being that even within the chips, where the narrative was most uniform, the highs are not synchronised.
In short: Filed under the credit-market side of the AI build-out: "Broadcom is reportedly seeking more than 60 billion in financing tied to its latest AI deal, which resulted in a huge spike in its CDS, with the full package potentially nearing 100 billion. The structure could include 60 to 70 billion of senior secured debt plus 30 billion of junior debt." The credit-default-swap move is the tell he points at, not the equity.
Broadcom designs custom AI chips for the largest technology companies. The item that caught Singh's eye is not a product but a financing: the company is reportedly seeking more than $60 billion of debt tied to its latest AI deal, with the total package potentially approaching $100 billion.
What matters is the market's reaction in the credit market rather than the stock market. Broadcom's credit default swaps — effectively the insurance premium other investors pay to protect themselves against the company failing to repay — spiked. That is bondholders repricing the risk of borrowing on this scale to fund equipment for a single customer relationship.
It belongs to the call's larger theme: the AI build-out has moved from being funded out of profits to being funded with borrowed money, and the credit market is beginning to charge for it.
Full passage: premium transcript (PDF).
In short: Held but explicitly not favoured. BMO initiates/reiterates outperform with a $455 target — 25% upside — and Brown uses the $1.7 trillion market cap in his roll call of AI-exposed size. Raskin owns it with NVIDIA and gives both the same lukewarm verdict: "we've owned them for ages… I'm not overweight either of the stocks. So we like them. But I think there are other more interesting stocks in the market at this point. They just have very high expectations in them. There's a lot of good news. The multiples are not rich. So we still own them — it's just getting to that next leg." A hold on expectations, not on valuation.
Broadcom designs the custom AI chips and networking silicon the large cloud companies buy, and at $1.7 trillion it is one of the biggest companies in the world. BMO has it at outperform with a target 25% higher.
Amy Raskin has owned it for years and her verdict is deliberately lukewarm: she is not overweight, she likes the company, but "there are other more interesting stocks in the market at this point." Her reason is not the price — she says the multiples are not rich — but the expectations embedded in the story: a lot of good news is already assumed, so the next move up requires something new.
That is a useful distinction. A stock can be reasonably valued and still be a poor use of capital, if what has to go right for it to work is already the consensus.
In short: Down on the Marvell-Alphabet headline, and Terranova argues the market is pricing the wrong risk: "I don't think Broadcom is down on the concerns that Marvell is going to take market share on the custom chips for the TPUs. It's more about, in the future, if Marvell builds the relationship with Alphabet, do they turn to Marvell for silicon purchases — that would be detrimental to Broadcom. So these relationships I think ultimately are going to continue." Simpson keeps it inside the bull case for Alphabet rather than against Broadcom: the TPU ecosystem now runs "Broadcom, NVIDIA and Marvell… it's just shoring up what they need."
Broadcom fell on the Marvell news, and Terranova thinks the market is worrying about the wrong thing. The immediate fear — that Marvell takes share in Alphabet's current custom AI chips (its "TPUs") — is not, in his view, what is being priced. The real question is the next round: once Marvell has a working relationship with Alphabet, does Alphabet start buying more of its silicon there instead?
That is a slower, more strategic risk than a lost contract, and it is why he expects the existing relationships to continue for now. Simpson's version of the same fact is bullish for the buyer rather than bearish for the supplier: Alphabet now has Broadcom, Nvidia and Marvell, and having three sources is exactly what a company spending this much on chips should want.
In short: One mention, and it is an exit: Druckenmiller "exited AVGO, INTC, and MU" in the same quarter he rebuilt Amazon and reopened Alphabet. Unlike INTC and MU, Broadcom draws no offsetting buy anywhere in the round-up. A disclosed author holding. (Caveat App Economy applies to all sells: hedge funds often trim high-conviction positions to manage risk, so a top sell "doesn't always mean a bearish turn.")
In short: Named inside the durable-demand list from the compute race — HBM, grid connection, liquid cooling, ASIC premiums "and you're going to see demand for Broadcom products" — set against "a symmetric risk of the cost of overbuilding."
21:46You're going to see demand for Broadcom products and eventually there's going to be a symmetric risk of the cost of overbuilding. So you're going to have a lot of unused GPUs because these data centers are going to take a long time to be completed. I mean they're running one to two-year delays. But for now, I think the hyperscalers are just focused on building as much compute as possible so that they can be the biggest cloud provider.
In short: Held "for years" — bought in 2019 as "a 4% yielding tech stock" before AI was in the story; his example of letting a winner run rather than selling on valuation.
Broadcom makes networking and custom chips plus infrastructure software. Hamlin bought it in 2019 as a boring 4%-yielding tech company, before anyone talked about AI, and has held it ever since as it became an AI winner. He uses it to show two things: dividend funds have little AI exposure, and you shouldn't sell a great business just because it now looks expensive.
21:45So they missed out on some of that. Some of the dividend ETFs were able to get their hands on Seagate or Qualcomm for a little bit. We own Broadcom, which we've owned for years, but it — when we bought that in 2019, it was a 4% yielding tech stock, and the words AI never came out of their mouth just yet.
In short: Link owns it and is "very pleased" — up 9% this week alone. Her framing pairs it with NVIDIA: both have badly lagged the semiconductor sector this year, "and that was the other reason why it was so interesting to me. They were such laggards and I thought they could play catch-up. Both have earnings power of $20 a share over time." The visibility argument: "Broadcom's AI semiconductor pipeline is $100 billion. There's a lot of visibility, they're doing a lot of great things."
Broadcom designs custom AI chips for the biggest cloud companies, and Stephanie Link owns it — up 9% this week. Her reason for holding both this and Nvidia is that both had badly lagged the semiconductor sector this year, which is unusual for the two most important names in it: "they were such laggards and I thought they could play catch-up. Both have earnings power of $20 a share over time." The visibility comes from a stated $100 billion pipeline of AI semiconductor work, which is what turns a cyclical chip company into something you can forecast.
In short: Baron (Global Durable Advantage ETF): a new position taken "advantage of the stock's sell-off post earnings." The moat claim is maximal — "one of the most durable and formidable competitive moats in the AI infrastructure buildout" — because hyperscalers optimising for "the highest intelligence per dollar of capex" must partner with "the best silicon design player." The bear case (customers vertically integrating) "remains only a bear narrative at this stage," since ASIC complexity "demands very tight integrations… extreme co-design of compute, memory, input/output on dies, and networking fabrics." Evidence: the Google TPU agreement extended to 2031 ("a commitment that is very substantial in dollars"), OpenAI's Jalapeno inference chip taped out "in a record nine months" targeting 10GW, plus Anthropic, Meta and a multi-year Apple ASIC deal. Hock Tan "expects ASICs to match GPU units in volume by next year."
Broadcom designs the custom AI chips that big technology companies use instead of buying everything from Nvidia — Google's TPUs are the best-known example — and it also sells the networking silicon and, through VMware, the software that manages large fleets of servers. Baron bought the shares into a post-earnings sell-off.
Their case is that the obvious bear argument — that customers will eventually design these chips themselves — underrates how hard the job has become. Modern AI systems require the compute, memory, chip-to-chip links and networking to be designed together, and Broadcom is at the front in all of them. A company whose revenue depends on having the best chips cannot afford to hand the work to a second-best in-house team.
The evidence they point to is contractual rather than rhetorical: Google extended its agreement with Broadcom to 2031, OpenAI got its first inference chip designed and taped out in nine months, and Apple signed a multi-year deal spanning several product generations. Broadcom's chief executive expects custom chips to match graphics-processor volumes next year.
Full passage: premium transcript (PDF).
In short: Market color: "nothing was wrong with Broadcom" either — cited (with Micron and ASML) as "arguably the first sort of rumble in the space," where a genuinely good report still couldn't clear "very lofty expectations." No individual call.
In short: Link: "Broadcom, I think they have $20 in earnings power" — a semi she likes within her barbell alongside NVIDIA and Marvell.
Broadcom makes networking chips and custom AI silicon for the big cloud companies. Stephanie Link's one-liner is that it has "$20 in earnings power" — i.e. she thinks its earnings can reach a level that makes today's price look cheap — so it's one of the chip names she likes to own alongside NVIDIA and Marvell.
In short: Two positive tells: it's Meta's design partner for the AI chip heading to production in September (TSMC-fabbed), and it rallied late in the week on an expanded ~$30B Apple deal (15B of US-made chips, a $1.5B Fort Collins plant).
Full passage: premium transcript (PDF).
In short: Giroux: diversity-of-compute winner; Ellenbogen notes the LLM companies have lessened Nvidia ties and worked more with Broadcom (plus Google TPUs and AWS Graviton). (Black's caveat: 32x — the other megacap he calls not reasonably priced.)
In short: Baruch is buying it alongside NVIDIA as he puts new money to work — "I'm trading the target, NVIDIA, Broadcom… these are buys right here and it's going to run."
Broadcom makes networking chips and custom AI silicon for the big cloud companies. Bill Baruch is buying it right alongside Nvidia as he deploys new cash — "these are buys right here and it's going to run" — treating the pair as his way to re-enter the AI-chip trade after its recent pause.
In short: Referenced as a former semiconductor leader now a laggard alongside NVIDIA — Stephanie Link made "a ton of money" in it over the years; fundamentals better, valuation pulled back.
In short: Her other semi — and the one she's been adding to. ~$20 of earnings power; the stock hasn't done as well as Marvell, so it's the relative laggard she's accumulating. Owns the custom-ASIC franchise (80% share with Marvell).
Broadcom is Link's second semiconductor holding and the one she's actively adding to. It makes the same kind of custom AI chips and networking gear as Marvell (the two dominate that custom-chip niche). She frames it on earnings power of about $20 a share, and notes the stock "hasn't done as well" as Marvell — so it's the relative laggard, which is exactly why she's buying more of it rather than the better-performed name.
In short: The most alarming link in "Big Sky": Broadcom gives a residual-value guarantee on the $30B A1/A2 tranches — i.e. it's on the hook for $30B if the GPUs/TPUs (it co-makes with Google) are worth too little. "Financing its own revenue"; "I frankly do not understand why they'd agree." Weak equity reflects the risk.
Broadcom is a giant chipmaker, but the worry here is a deal, not its products. To help fund the AI startup Anthropic, Apollo and Blackstone raised $35 billion by selling bonds backed by AI chips. Because banks won't lend against chips alone (they lose value fast), Broadcom agreed to guarantee the senior $30 billion — meaning if Anthropic can't pay and the used chips don't fetch enough, Broadcom eats the shortfall.
The circular part: that borrowed money buys chips Broadcom helps make (with Google, which also owns 14% of Anthropic) — so Broadcom is effectively guaranteeing loans that pump up its own sales, and the same bank (Morgan Stanley) advised Broadcom and lent money to the bond buyers, who weren't even shown Anthropic's financials. Singh can't understand why Broadcom would take that $30 billion risk, and reads the whole structure as a flashing sign of an overheated AI-financing bubble.
Full passage: premium transcript (PDF).
In short: Reported AI numbers "while strong were below expectations," which sent the semiconductor group into a correction Thursday — one of the two triggers (with the 10-year) for Friday's selloff.
2:22Big moves. What happened started Wednesday night when Broadcom reported AI related numbers that while strong were below expectations. As a result, on Thursday, the semiconductor group experienced a correction. On Friday, the big news was that the employment numbers were better than expected. Actually, they were quite strong, implying the Fed will not be raising rates and might even raise them.
In short: Celestica's close partner on the open-source SONiC switching software — central to AI networking (referenced, not a standalone call here).
53:15So what we thought was just a great growth driver turned out to be great competitive advantages and they have like 50 60% share of the cloud Ethernet switch market which is a crucial market for um AI because AI is incredibly network intensive. And then even something like the printed circuit board. I mean a regular server you need 10 layers.
In short: The trigger for last week's crack — one of the AI trade's biggest winners delivered a beat but no raise (just reiterated), a deceleration vs analyst expectations. Down ~17% on the week; "not a death blow, but this wasn't good."
3:12And investors were expecting the same thing that we've seen over and over again, a beat and a raise. You have to have that combination to have these incredible performances that we've seen. But Broadcom did not give investors a beat and a raise. They basically just beat and then they reiterated the same thing that they said last time. That's a change in pace.
In short: AI business (custom chips — e.g. Google's TPUs, a 15-year relationship — plus AI networking) guided to $100B next year, "way bigger than the entire company was a year or two ago," and "they'll probably do a lot better than 100." Stock lagged like Nvidia (accelerator fatigue + the software taint), but its software (VMware) is infrastructure AI runs on — a baby thrown out with the SaaS bathwater.
Broadcom is two companies in one: a sprawling chip business and a software business (VMware and others). The part that matters now is AI: Broadcom designs "custom" AI chips for the giants — most famously Google's TPUs, a partnership running 15 years — plus the networking chips that tie thousands of AI processors together. That AI business is forecast to do $100 billion of revenue next year, more than the entire company's sales a year or two ago, and Rasgon's guess is "they'll probably do a lot better than 100."
The stock has lagged for the same reasons Nvidia has (investors chasing scarcer parts of the supply chain), plus guilt by association with software stocks, which the market has been dumping on fears AI will replace them. Rasgon calls that a mistake here: Broadcom's software is plumbing (the "virtualization" layer that cloud computing — including AI — actually runs on), not the kind of application AI threatens. A baby thrown out with the bathwater.
41:52So, and they've they've been doing these chips, by the way, for 15 years. It just it wasn't that big until fairly recently. They've been working with Google for for 15 years, but now with AI, it's just taken off. And so, this overall AI business across the uh the custom chips and and the AI networking, they guided for next year for that to be a hundred billion dollars, which is way bigger than the entire company was, you know, a year or two ago.
In short: Blow-out Q2: revenue +48% Y/Y to $22.2B, AI semis +143% to $10.8B, record 49% op margin, FCF $10.3B. The tell was order velocity — AI bookings >$30B, ~3× what it shipped, CEO Hock Tan: "visibility runs all the way to 2028" ($100B FY27 AI target, OpenAI 10 GW / Meta 3 GW deals). Yet shares fell ~15% — the target was reiterated not raised, and Tan acknowledged Google will diversify TPU supply. Strong fundamentals, cautious tape.
Broadcom designs custom AI chips ("XPUs") and the networking gear that wires data centers together. The most important number this quarter wasn't revenue (which jumped 48%) — it was bookings: orders for $30 billion of AI chips, about three times what Broadcom actually shipped. That ratio (bookings divided by shipments, sometimes called "book-to-bill") is a forward-demand tell: when orders pile up far faster than you can deliver, customers are signaling years of demand ahead. The CEO put it bluntly: "visibility runs all the way to 2028." Those orders come from OpenAI (a 10-gigawatt deal), Meta (3 GW), and an Anthropic chip purchase Broadcom helped finance.
So why did the stock drop 15% on a blow-out? This is the classic "great quarter, disappointing guide" trap. The market had bid the stock up expecting Broadcom to raise its $100 billion AI-revenue target for 2027; instead it merely repeated it. And the CEO admitted Google — one of Broadcom's biggest custom-chip customers — will start buying some of those chips elsewhere too. Strong business, but the bar was set higher than even great results could clear. App Economy's read leans positive on the fundamentals while flagging that the tape stayed cautious.
In short: The June-5 sell-off trigger — the Broadcom selloff was "pulling down the entire tech sector."
3:00I mean this is a yield breakout in almost all the developed worlds as you know. So, it's it's a continuation of what has been a multi-year global bond bare market for the rich count so-called rich countries. — Stock markets are taking a huge hit today on the 5th of June. Now, some of this has to do with the Broadcom selloff that's pulling down the entire tech sector. The S&P is down 2 and a half%.
In short: With Intel & AMD, has "lost some momentum" — technical evidence the AI rally's leadership is thinning.
Broadcom is a key AI-chip supplier. Woo groups it with Intel and AMD as names that have "lost momentum" — once-leading chip stocks that have stalled. The point is that the AI rally's leadership is thinning out, reinforcing his view that the trade is crowded and vulnerable.
10:02This could show up in the earnings and guidance for Q2. I see few positive fundamental catalysts for the AI trade in the near term. And technically, Intel, AMD, and Broadcom all seem to have lost some momentum. What this means is that the AI rally might become even more narrow. I think the balance of risk is starting to favor the downside as the AI trade is getting very crowded.
In short: A networking seller he thinks is "becoming less cyclical over time" and "probably will do well" — a valid phase-1 pick, but not one of his big bets.
8:18ASML still have a little bit of cyclicality. Then, of course, we have TSMC. TSMC is a very durable seller. It's not some highly cyclical stock. We have Nvidia which I believe because of CUDA, because of the infrastructure, because of the ecosystem, it has some cyclicality, but I truly believe that Nvidia is becoming less cyclical over time.
In short: "I would never bet against Broadcom" (Google's TPU design partner) — a respectful aside, not a direct pick in this conversation.
18:28But listen, the TPU is a great chip. I'm sure the TPU V9 is going to be amazing. They'll make more aggressive choices. I would never bet against Google. I would never bet against Broadcom. But I do think Trainium is super underestimated right now. Yeah, I appreciate it. I want to switch gears to a topic which actually is how we first got connected way back in 2022 and then 2023.
Nothing matches this filter.
Verbatim excerpts from the public transcripts (auto-pulled at each mention's timestamp, lightly cleaned). Timestamps link into the video; "source page" opens that commentator's full analysis at this row. Click a mention's header line to expand it (one open at a time). For personal study — not investment advice.