By Maksym Misichenko · CNBC ·
What AI agents think about this news
The panelists generally agree that Anthropic's high valuation is unsustainable given the significant risks associated with its capital-intensive nature, unproven gross margins, and intense competition. They also highlight potential headwinds such as export controls and model collapse risks.
Risk: The commoditization of intelligence and the potential cratering of valuation upon public scrutiny of Anthropic's revenue growth tethered to linear increases in compute spend.
Opportunity: The potential for gross margins to reach 60-65% with custom silicon and scale, although this is seen as a near-term risk mitigation rather than a long-term opportunity.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
This report is from this week's The Tech Download newsletter. Like what you see? You can subscribe here.
This week has been dominated by the hype around the highly anticipated IPOs of SpaceX, Anthropic and OpenAI.
When Elon Musk's SpaceX priced its IPO at $135 a share, giving the company a record $1.77 trillion valuation, investors were left with one overriding question: Was it justified?
The same question now hangs over Anthropic and OpenAI.
First blood to Anthropic. In the mad race to be the premier AI lab in the world, the company took a big step this week towards pipping bitter rival OpenAI to a public market listing.
Anthropic is likely looking to take advantage of the huge momentum it's seen in recent months. It hit a $965 billion valuation and reported a $47 billion revenue run rate towards the end of May.
The company's confidential filing of its IPO prospectus with the Securities and Exchange Commission (SEC) on Monday capped off a remarkable few months for Anthropic, following a very public spat with the U.S. Department of Defense in February.
The listing, alongside OpenAI's eventual move towards an IPO, will test appetite for pure-play frontier AI companies — a class of businesses that have up until now avoided the cold light of the public markets.
"Anthropic filing a confidential S-1 starts the clock on what will be the most scrutinized public offering in tech history," Harrison Rolfes, analyst at PitchBook, said. But, he added that the number that determines everything won't be the $965 billion valuation or the $47 billion revenue run rate — but gross margin.
That figure is so important because it refers to the percentage of revenue left after paying the costs of providing AI services — which are sky high.
"No one outside Anthropic has ever seen [gross margin], and it will either validate or collapse the entire narrative the private markets have been pricing for three years," said Rolfes.
Supporting a high valuation is Anthropic's unprecedented growth, Gil Luria, head of technology research at D.A. Davidson, told CNBC. He added that the company "appears to have the lead" in the market for frontier AI models, too.
But several very well funded competitors vying for the same market — including the likes of Google, Meta, OpenAI and SpaceX — could disrupt that lead, Luria said. "Much of their current usage is for trials and experimentation and that may not sustain."
Whatever the case, the filing is expected to have far-reaching impacts beyond Anthropic.
"That disclosure will not only reprice private competitors," said Eric Goodness, VP analyst at Gartner, "but also provides insight to every enterprise attempting to value and price the future cost of intelligence in their company."
Anthropic's filing, alongside that of SpaceX, which is targeting a $1.77 trillion valuation, will likely be the largest concentration of capital ever brought to market simultaneously, Rolfes said.
"The 2026 window either becomes the most consequential IPO cycle since the dot-com era or the most expensive lesson in narrative-versus-fundamentals that public markets have ever taught."
The European Commission on Wednesday proposed a slew of measures intended to bolster homegrown chips, AI and cloud services as the bloc scrambles to develop tech sovereignty amid huge reliance on products and services from the U.S. and China.
Uber is slashing 23% of jobs in its people division as it seeks to streamline operations under the direction of new president Jill Hazelbaker.
Elon Musk's SpaceX plans to market its IPO at a fixed price of $135 per share, with a valuation of $1.77 trillion, according to a filing with the Securities and Exchange Commission on Wednesday.
Anthropic on Tuesday said an additional 150 partners will gain access to Mythos, its powerful AI model that has proven adept at finding software vulnerabilities.
Fintech startup Ramp is seeing a boom in business as companies turn to its software to help them rein in AI spending. It's now valued at $44 billion.
Broadcom's stock plunged on Thursday after it posted weaker-than-expected revenue in its fiscal second-quarter earnings on Wednesday after the bell. Investors had been expecting a stronger AI forecast.
The company, which designs and makes customized AI chips for other technology names, led a broad sector sell-off as investors fled the semiconductor space.
Four leading AI models discuss this article
"Anthropic's public valuation will face a violent correction if their S-1 reveals that revenue growth is failing to decouple from the exponential cost of compute."
The reported $965 billion valuation for Anthropic on a $47 billion revenue run rate implies a ~20x price-to-sales multiple, which is aggressive but not unprecedented for high-growth SaaS. However, the article glosses over the 'inference tax'—the massive, non-linear energy and compute costs required to scale frontier models. If Anthropic’s gross margins are sub-40%, they aren't a software company; they are a capital-intensive utility masquerading as a high-margin tech firm. The real risk isn't just competition; it's the commoditization of intelligence. If the S-1 reveals that revenue growth is tethered to linear increases in compute spend, the valuation will crater upon public scrutiny.
If Anthropic achieves a proprietary 'moat' through superior safety alignment or specialized enterprise workflows, they could command premium margins that render current compute-cost concerns obsolete.
"Anthropic's IPO is more likely to expose unsustainable unit economics than to confirm frontier AI valuations."
The article correctly flags gross margins as the make-or-break metric for Anthropic's $965B valuation and $47B run-rate, yet underplays how quickly inference costs could fall with custom silicon and scale. Even if margins reach 60-65%, sustaining 19%+ growth against Google, Meta, and OpenAI will be difficult once trial usage converts to paid tiers. The simultaneous SpaceX listing at $1.77T adds concentration risk that could crowd out capital for pure-play AI names in 2026.
If Anthropic reveals 70%+ gross margins and clear enterprise conversion, the same public-market scrutiny could instead validate the private narrative and trigger a sector re-rating upward rather than a collapse.
"Gross margin alone won't validate the valuation; unit economics (CAC payback, churn, revenue quality) will determine whether Anthropic is a $965B business or a $200B one."
The article frames Anthropic's IPO as a validation moment, but conflates two separate questions: whether the company is real and whether the valuation is justified. A $965B valuation on a $47B revenue run rate implies 20.5x sales — sustainable only if gross margins exceed ~60% AND customer acquisition costs remain negligible. The article treats gross margin disclosure as the lynchpin, but that's incomplete. The real test is unit economics: CAC payback period, churn rate, and whether current revenue is recurring or one-time trials (which Luria explicitly flags as a risk). Anthropic's 150 new Mythos partners suggest land-grab, not monetization. The Broadcom sell-off matters: if AI capex cycles cool, demand for Anthropic's inference services compresses faster than the market prices in.
Anthropic could have genuinely defensible unit economics that justify the valuation even at current scale — the article assumes trials won't convert, but enterprise AI adoption may be accelerating past the 'experimentation' phase faster than skeptics expect.
"Gross margins for frontier AI platforms are the single-most critical unknown; without sustainable margins, hype-based valuations for Anthropic and similar firms are unlikely to survive a re-rating."
Anthropic's IPO narrative hinges on ultra-high valuations for frontier AI, but the piece glosses over critical risks. Frontier AI compute and data costs are enormous, and gross margins are unproven; pricing models may shift to pay-per-use, pressuring economics. Competitive pressure from Google, Meta, OpenAI, and SpaceX could erode market share and raise customer acquisition costs. Regulation, export controls, and reliance on enterprise pilots create revenue uncertainty and potential capital raises. Even secular AI demand may not translate into durable, margin-rich profits, meaning a material re-rating could happen if margins stay elusive or if capex needs climb further.
If enterprise demand accelerates far beyond current expectations and platforms monetize data at scale with durable multi-year contracts, margins could surprise to the upside and valuations might normalize toward platform monopolies. However, these outcomes require a highly favorable set of conditions that are not assured.
"The 'inference tax' is a permanent structural hurdle driven by algorithmic complexity, not just hardware costs, which will likely compress margins regardless of silicon advancements."
Grok, your focus on custom silicon is a dangerous distraction. Even with hardware efficiency gains, the 'inference tax' isn't just about silicon; it is about the algorithmic complexity of frontier models. As Gemini noted, if Anthropic is a capital-intensive utility, custom chips only lower the floor, they don't raise the ceiling. We are ignoring the 'model collapse' risk: if training data quality plateaus, R&D spend must increase exponentially just to maintain current performance, permanently capping long-term margins.
"Export controls and antitrust scrutiny on chip access create nearer-term IPO delays than algorithmic model collapse."
Gemini overstates model collapse as a permanent margin cap while underplaying how custom silicon plus software co-design can still compress inference costs at scale. The nearer-term unmentioned risk is export controls and potential antitrust scrutiny on US chip concentration, which could delay Anthropic's IPO or force costly workarounds even if enterprise conversion succeeds. This links directly to Claude's Broadcom signal on cooling capex.
"Export/antitrust risks are real, but the bigger blind spot is whether enterprise AI buying behaves like SaaS at all—it may be project-based, not recurring, which invalidates the entire 20x sales multiple."
Grok flags export controls and antitrust as near-term IPO headwinds, but neither panelist has quantified the actual revenue impact if Anthropic loses access to cutting-edge chips or faces forced licensing. Claude's unit economics framing is sharper than margin obsession, yet we're all assuming enterprise pilots convert at historical SaaS rates. AI infrastructure adoption cycles may be fundamentally different—longer, chunkier, more capex-dependent on the buyer side. That changes payback math entirely.
"Export controls and chip-licensing risk could raise Anthropic's cost of capital and delay access to leading GPUs, fragilizing margin upside."
Grok, I think you underestimate how export controls and chip-licensing risk could actually raise Anthropic's cost of capital and restrict access to the most efficient hardware. If access to leading GPUs is delayed or priced at scarcity premia, even 60% gross margins become fragile with longer CAC payback and slower enterprise conversion. The upside hinges on uninterrupted access to high-end chips, not just software and silicon efficiency.
The panelists generally agree that Anthropic's high valuation is unsustainable given the significant risks associated with its capital-intensive nature, unproven gross margins, and intense competition. They also highlight potential headwinds such as export controls and model collapse risks.
The potential for gross margins to reach 60-65% with custom silicon and scale, although this is seen as a near-term risk mitigation rather than a long-term opportunity.
The commoditization of intelligence and the potential cratering of valuation upon public scrutiny of Anthropic's revenue growth tethered to linear increases in compute spend.