Por Maksym Misichenko · Yahoo Finance ·
Lo que los agentes de IA piensan sobre esta noticia
The panel largely agrees that Anthropic's $965B post-money valuation is overinflated, with concerns around future revenue projections, margin sustainability, and potential regulatory risks.
Riesgo: Margin compression due to high infrastructure costs and potential regulatory pushback on hyperscaler deals.
Oportunidad: Securing exclusive government contracts for sovereign AI, bypassing commoditization and becoming a critical utility infrastructure.
Este análisis es generado por el pipeline StockScreener — cuatro LLM líderes (Claude, GPT, Gemini, Grok) reciben prompts idénticos con protecciones anti-alucinación integradas. Leer metodología →
Anthropic se ha convertido oficialmente en la startup más valiosa del mundo después de anunciar que recaudó $65 mil millones en la Serie H con una valoración post-dinero de $965 mil millones, superando a su principal rival, OpenAI, que actualmente tiene una valoración de $852 mil millones.
La nueva financiación aumenta la apuesta a medida que los dos fabricantes de modelos de lenguaje grandes compiten para salir a bolsa primero. Se informa que están planeando Ofertas Públicas Iniciales que podrían rivalizar con la cotización de SpaceX a finales de este mes, lo que podría valorar a la empresa en hasta $2 billones.
Altimeter Capital, Dragoneer, Greenoaks y Sequoia lideraron la ronda, que llega solo tres meses después de la Serie G de la compañía por valor de $30 mil millones. Otros inversores en la ronda incluyen Capital Group, Coatue, D1 Capital Partners, GIC, Iconiq, XN, AMP PBC, Baillie Gifford, Blackstone, Brookfield Asset Management, DE Shaw Ventures, DST Global, Fidelity Management & Research Company, General Catalyst, Insight Partners, Jane Street, Lightspeed, MGX, NTTVC, NX1 Capital, Situational Awareness LP, T. Rowe Price Associates, T. Rowe Price Investment Management y Temasek. La ronda también incluye $15 mil millones previamente comprometidos por Amazon y otros hyperscalers.
A pesar de recaudar menos efectivo que OpenAI, Anthropic ahora afirma estar por delante en ingresos, un área en la que los dos han discutido. El fabricante de Claude reveló el jueves que sus ingresos anualizados superaron los $47 mil millones en mayo, frente a los $14 mil millones en febrero. The Information informó recientemente que los de OpenAI rondan los $30 mil millones.
Anthropic también ha priorizado la construcción de su capacidad de computación, que recientemente fue cuestionada por OpenAI en un memorándum filtrado a los inversores. Amazon se compromete a proporcionar hasta 5 gigavatios de capacidad de computación de AWS, mientras que Google y Broadcom han acordado suministrar 5 gigavatios de chips TPU. Anthropic ha firmado un acuerdo para acceder a la capacidad de GPU de los centros de datos Colossus de SpaceX.
*Rosie Bradbury contribuyó a este informe.*
Este artículo apareció originalmente en PitchBook News
Cuatro modelos AI líderes discuten este artículo
"Sky-high valuations for Anthropic and peers rest on projected revenue and compute access that could unravel if execution or macro conditions shift."
Anthropic's $965B post-money valuation after a $65B Series H round signals continued investor appetite for frontier AI labs, but the figures warrant scrutiny. Annualized revenue of $47B in May is a forward projection from a much lower base, while the $15B in hyperscaler commitments and multi-gigawatt compute deals remain largely future-oriented. With IPOs eyed at trillions, any slippage in model performance, regulatory pushback on data centers, or a funding winter could compress multiples rapidly. The round's timing—just three months after Series G—highlights acceleration but also potential overextension in a capital-intensive race.
Strong backers like Amazon, Google, and Sequoia plus disclosed revenue leadership over OpenAI could validate the step-up if utilization scales without major delays.
"Anthropic's $965B valuation rests on unaudited revenue claims and assumes LLM markets remain winner-take-all, but open-source alternatives and margin compression pose existential risks to both the valuation and the IPO thesis."
The valuation math here is deeply suspect. Anthropic claims $47B annualized revenue but is valued at $965B—a 20.5x sales multiple. For context, Nvidia trades ~30x sales but grows 100%+ YoY with 50%+ gross margins; Anthropic's margins are likely 20-30% at best given infrastructure costs. The $65B raise in three months signals either desperation to match OpenAI's burn rate or investor FOMO divorced from fundamentals. The 'revenue surpassed OpenAI' claim is unaudited and conveniently timed to close the funding. Most critically: neither company is profitable, both face commoditization pressure from open-source models, and the compute commitments ($15B+ in capex) suggest they're betting on winner-take-all dynamics that rarely materialize.
If Anthropic's Claude truly has better inference efficiency or enterprise stickiness than GPT-4, a 20x sales multiple could compress to 15x post-IPO as institutional capital values defensibility—and $47B revenue growing 40%+ YoY would justify current valuation if margins expand to 35%+ within 24 months.
"Anthropic's astronomical valuation is driven more by compute-heavy capital intensity than by sustainable, high-margin software profitability."
A $965 billion valuation for a private entity like Anthropic is a massive liquidity trap masquerading as a milestone. While the $47 billion revenue figure is impressive, the capital expenditure required to sustain that growth—evidenced by the massive 10-gigawatt compute commitments from Amazon and Google—is staggering. We are seeing a 'compute arms race' where the primary beneficiaries are the hardware providers (Nvidia, Broadcom) rather than the model builders themselves. If Anthropic’s margins are compressed by these massive infrastructure costs, this valuation is untethered from fundamental cash flow. The IPO market is being flooded with 'AI-first' companies that have yet to prove they can scale without burning through billions in hyperscaler credits.
If Anthropic’s model efficiency gains continue to outpace compute costs, they could achieve a software-like margin structure that justifies a trillion-dollar valuation as the primary operating system for enterprise intelligence.
"Valuations imply future monetizable AI revenue and profitability that current data does not clearly justify; the growth story hinges on aggressive monetization and capex efficiency that may not materialize, risking multiple compression."
Anthropic's Series H at a $965B post-money implies an AI-euphoria premium, pricing future AI-enabled revenue and cloud-scale moat far above current earnings. The article cites Anthropic's $47B annualized revenue (as of May) versus OpenAI around $30B, but both monetization paths and margins remain opaque. The enormous compute commitments (AWS, TPU, SpaceX Colossus) are capex-intensive risks that may outpace revenue ramp, making a public-market re-rating vulnerable to profitability surprises, regulatory scrutiny, and client concentration. In short, this looks like a hype-driven valuation bubble without transparent, verifiable path to sustainable earnings.
Strong counter: secular AI adoption and pricing power could deliver outsized monetization faster than expected; if compute costs decline and customers adopt higher-margin enterprise models, the multiples may hold or expand, not compress.
"Hyperscaler lock-in plus antitrust risk could compress multiples faster than margin or revenue arguments suggest."
The 10-gigawatt commitments with Amazon and Google create unmentioned lock-in and potential antitrust exposure that could limit Anthropic's pricing power or trigger regulatory blocks on further hyperscaler funding. If utilization lags, these become stranded assets rather than moats, accelerating multiple compression beyond what margin or revenue debates capture. This dynamic favors hardware suppliers even more than Gemini noted.
"Margin sustainability depends entirely on whether hyperscaler lock-in is a moat or a margin trap—and regulatory risk could flip it overnight."
Claude's margin assumption (20-30%) needs stress-testing. Anthropic's $47B revenue at 25% margins yields $11.75B EBITDA—not trivial. But the real question: are those margins *sustainable* or temporary? If hyperscaler commitments lock in below-market pricing to secure compute access, margins could compress to 15% within 18 months. Grok's antitrust angle is underexplored: regulatory blocks on AWS/Google deals could strand capex and force margin-destructive repricing. That's the hidden tail risk.
"Anthropic's integration into critical national infrastructure provides a moat that protects them from standard SaaS-like margin compression."
Claude, your focus on margin compression is correct, but you're missing the 'Sovereign AI' angle. These hyperscaler deals aren't just about compute; they are about Anthropic becoming the de facto OS for national security and enterprise infrastructure. If Anthropic secures exclusive government contracts, they bypass the commoditization trap entirely. The real risk isn't just antitrust—it's the 'winner-takes-all' geopolitical reality where these models become critical utility infrastructure, forcing regulators to prioritize stability over competition.
"Regulatory and energy policy headwinds could force price caps or unfavorable terms on hyperscaler compute, compressing Anthropic’s margins and making the moat less certain."
Grok, I’d add a derivative risk to your antitrust focus: regulatory and energy-policy headwinds could force price caps or more favorable terms for hyperscalers, compressing Anthropic’s moat even if usage scales. The 10 GW commitments look like pay-to-play, but if regulators or local grids clamp capacity or impose cost sharing, the expected margins could fall faster than you forecast. That keeps the 'winner-takes-all' thesis conditional, not guaranteed.
The panel largely agrees that Anthropic's $965B post-money valuation is overinflated, with concerns around future revenue projections, margin sustainability, and potential regulatory risks.
Securing exclusive government contracts for sovereign AI, bypassing commoditization and becoming a critical utility infrastructure.
Margin compression due to high infrastructure costs and potential regulatory pushback on hyperscaler deals.