CNBC Squawk on the Street with Pauline Yang - May 29th, 2026
An 11-minute CNBC Squawk on the Street exclusive with **Pauline Yang, the Altimeter partner who led Anthropic's Series H** — *'now Altimeter's biggest investment to date.'* The headline: the round prices Anthropic at a **$965B post-money valuation** (alongside Dragoneer, Green Oaks and Sequoia), up from **$380B in February** — a sprint to *'almost a trillion'* in roughly four months. Yang's job here is to defuse the bubble read, and her core argument is a multiple, not a vibe: at the **$380B round the company was at ~$9B run-rate revenue; today it announced $47B run-rate**, so *'the revenue multiple for this round is actually cheaper than the last round, sitting around 20x.'* The growth (a **5x in run-rate in the first few months**, explicitly **not** her base case for the rest of the year) is doing the work. On the **IPO**: the *'rumored IPO timeline is October'* and the raise is opportunistic — *'if you can raise, go raise… more cash and a stronger balance sheet never hurts.'* On the **OpenAI rivalry** she refuses to pick a sole winner — the TAM is *'the largest… we've ever seen in multiple generations'* and, like internet/mobile/AI-infra super-cycles that *'each generated a $5 trillion company,'* both can be *'multi-trillion-dollar winners,'* though she concedes Anthropic *'has executed with their product velocity.'* The most load-bearing defensibility claim answers the commoditisation bear directly: models *'can be swapped in or out,'* but **the harness — the product on top, plus integrations (Salesforce, HubSpot) and context — drives stickiness**, backed by a CFO data point of **net retention over 500%**. On margins, she leans on **$47B run-rate against ~3,000 employees** — *'that is just unheard of.'* And she is *'very, very bullish on Nvidia'* (the stock *'up 16% for the year,'* now *'a key partner for Anthropic'*), flagging a persistent compute shortage and an Anthropic–SpaceX deal to lease **Colossus 1 and Colossus** capacity. This is the **bull-side bookend to last week's Brad Gerstner Altimeter hit** (Issue 07) and the same-house counterweight to the cohort's *'100x sales'* IPO anxiety.
Key points
- **The whole bull case is one number: a 20x revenue multiple, lower than the last round.** *'It's a big headline number, but… at the $380 billion round the company was sitting around $9 billion of run-rate revenue. And… today at 965, the company announced that they are at $47 billion of run-rate revenue. So actually the revenue multiple for this round is actually cheaper than the last round, sitting around 20x.'* This is the precise rebuttal to the [20VC roundtable's '100x trailing sales… I wouldn't buy a share' read on the IPO cohort (Issue 07)](/issues/2026-05-31): the lead investor's defence is that the denominator grew faster than the price.
- **The valuation moved $380B → $965B in ~4 months because run-rate 5x'd — and she is explicit it won't repeat.** *'the first few months the company has 5x in run rate revenue. Certainly we are not projecting that through the rest of the year, but we do think that the company will continue to add incremental ARR every month.'* A useful calibration flag from the buyer herself: the re-rate was a one-off step-change, not an annualisable run-rate. This is the same Anthropic that [lapped OpenAI at $44B ARR in Issue 07](/issues/2026-05-31), now marked at $47B and a near-trillion valuation.
- **On the IPO: the raise is opportunistic, the timeline is October.** *'The rumored IPO timeline is on October… investors would tell the company that if you can raise, go raise… more cash and a stronger balance sheet never hurts.'* Anthropic is raising privately on the brink of a public listing because demand is there — directly feeding [this week's $4T AI-IPO-wave thread (Laffont) and the OpenAI/SpaceX S1 supply wave](/issues/2026-06-07).
- **She won't crown a sole winner — internet/mobile/AI-infra each minted a $5T company, so both Anthropic and OpenAI can.** *'This is the largest TAM and market that we've ever seen in multiple generations… the Internet, the mobile, now AI infra — those have each generated a $5 trillion company. So we absolutely believe that both Anthropic and OpenAI can be multi-trillion-dollar winners over the course of time.'* Note the conflict-of-interest tell she discloses: Altimeter holds **both** Anthropic (now its largest) and OpenAI — so 'both win' is also her book talking.
- **The defensibility answer to the commoditisation bear: it's the harness, not the model.** *'Models certainly can be swapped in or out. But… the harness — the product that sits on top of the models — drives stickiness… the integrations with Salesforce or HubSpot… and the habit and the context that these models have… drives incredible stickiness.'* This is the direct counter to [the Mercor 'application-layer companies have no defensibility' thread in this week's cohort](/issues/2026-06-07) and to Issue 07's model-interchangeability worry — Altimeter's bet is that switching costs live in the workflow layer, not the weights.
- **The stickiness claim has a number behind it: net retention over 500%.** *'Anthropic's Chief Financial Officer provided this data point that the company's net retention rate is over 500%.'* She frames AI spend as following the cloud's expansion/optimisation cycle and insists *'we're still in an expansionary period'* — the bull-side mirror of [Issue 07's ROI reckoning](/issues/2026-05-31), where Microsoft cut Claude licences and a Fortune-20 spent $200M on tokens with 'minimal results'. Same fact, opposite spin: the question is whether 500% NRR survives the measurability gate.
- **The margin story is operating leverage: $47B run-rate on ~3,000 employees.** *'47 billion run-rate with 3,000 employees or so. That is efficiency. That is just unheard of.'* Plus gross-margin tailwinds as *'inference costs continue to come down… across a heterogeneous base of chips… and algorithmic improvements.'* This is the quantified version of [Issue 07's margin-ramp thesis (38% → 70% gross margin, $559M projected Q2 operating profit)](/issues/2026-05-31).
- **Very, very bullish on Nvidia — and it's now 'a key partner for Anthropic.'** Asked if Nvidia (*'up 16% for the year… is it mispriced?'*) is the laggard, Yang: *'we're very, very bullish on Nvidia… the compute shortage continues to be a story that both Anthropic and Open[AI] have to wrestle with… inference is going to continue to grow on this exponential… and Nvidia is very, very well positioned.'* A named long that fits squarely inside [Loeb's Nvidia/Anthropic/'Elon World' triad and the SOX regime change from Issue 07](/issues/2026-05-31).
- **The compute-supply tell: Anthropic is leasing SpaceX's Colossus capacity.** *'Anthropic did a deal with Space X to lease both Colossus 1 and Colossus to compute… a very strong signal that the company is certainly able to take on more compute and deploy it efficiently.'* This is the Anthropic side of [the SpaceX 'turning electrons into tokens' compute-deal narrative Gerstner pushed last week (Issue 07)](/issues/2026-05-31) — and the demand-side confirmation of the memory/compute-shortage trade.
- **The buried risk she was forced to address: a US-government 'supply chain risk' designation on Anthropic.** *'Certainly that's a risk. But… particularly with… the Mythos model, the administration has been much more open to working with… Anthropic… the Department of War will work with other models, but… the White House will continue to be a key partner.'* The one genuine bear item on the tape — a regulatory/procurement overhang that could matter 'down the road if they don't win in court' — which she frames as manageable given the TAM. Worth holding against [Issue 07's open-weight-ban / AI-legitimacy regulatory-tail thread](/issues/2026-05-31).
Notable quotes
It's a big headline number but I think what's missing is kind of if you look at the revenue growth that has happened from the beginning of the year till today when the $380 billion round happened, which we were also a part of, the company was sitting around 9 billion of run rate revenue. And if you look at the same valuation today at 965, the company announced that they are at 47 billion of run rate revenue. So actually the revenue multiple for this round is actually cheaper than the last round sitting around 20x.
the rumored IPO timeline is on October, I think certainly investors would, would tell the company that if you can raise, go raise. And while you know the company, I think there's a lot of operating leverage in the company that the company is seeing. Certainly more cash and a stronger balance sheet never hurts.
First and foremost, we think this is the largest TAM and market that we've ever seen in multiple generations. And so certainly we absolutely believe that there can be multiple win winners. If you look at the super cycles over the last few years, the Internet, the mobile, now air infra, those have each generated a $5 trillion company. So we absolutely believe that both Anthropic and OpenAI can be multi trillion dollar winners over the course of time.
We strongly believe that we're still in an expansionary period. Anthropic Chief Financial officer provided this data point that the company's net retention rate is over 500%.
models certainly can be swapped in or out. But I think what we've seen over the last six months is that the harness, right, the product that sits on top of the models drive stickiness. Right. If you think about the workflows that are being automated, the enterprises, if you think about the integrations, right, the integrations with Salesforce or HubSpot or whatever the connectors are and the habit in the context that these models have, I think all of that drives incredible stickiness amongst the enterprise.
we're very, very bullish on Nvidia. If you think about kind of the, what Anthropic has been really focused on in the last six months, it's getting more compute. Right. Whether that's through Amazon and through Google and Nvidia. Right. Nvidia is now a key partner for Anthropic.
the first few months the company has 5x in run rate revenue. Certainly we are not projecting that through the rest of the year, but we do think that the company will continue to add incremental ARR every month and continue to grow.
47 billion run rate with 3,000 employees or so. That is efficiency. That is just unheard of. And so, you know, we think that this company continue to drive very attractive margins, both on the gross margin and the operating margin side.
you saw that Anthropic did a deal with Space X to lease both Colossus 1 and Colossus to compute. And so I think that's a very strong signal that the company is certainly able to take on more compute and deploy it efficiently to customers.
Themes
- Anthropic valuation & revenue-multiple defence
- AI IPO supply wave
- Model commoditisation vs harness stickiness
- AI compute shortage & Nvidia
- Enterprise AI ROI / net-retention durability
Mentioned
Companies
Ideas
- Series H at $965B post-money
- $380B → $965B in ~4 months
- cheaper multiple at 20x revenue
- $9B → $47B run-rate (5x)
- October IPO timeline
- multi-trillion multiple-winners / $5T super-cycles
- harness stickiness over model commoditisation
- net retention over 500%
- operating leverage / $47B on ~3,000 employees
- Nvidia compute-shortage long
- Anthropic–SpaceX Colossus lease
- supply-chain-risk government designation