Open-weight and small models dominate as AI workloads shift to cheaper options
A healthcare technology company is shifting most AI workloads to open-weight and small models, achieving major cost savings while maintaining performance.
Investors reportedly weighed a $1.5 trillion valuation for Anthropic’s November IPO, down from $2 trillion expectations after a 45% revenue jump disappointed.
Anthropic’s IPO could come with a valuation about $500 billion lighter than investors once expected.
Some current and prospective investors are said to have talked over a roughly $1.5 trillion valuation for the November listing.
Reuters said the lower figure emerged from investors at Anthropic’s investor day in San Francisco on Wednesday. The AI company described the quarter as profitable on an adjusted basis.
Third-quarter revenue came to $16.8 billion, up 45%. The slower pace still left some investors dissatisfied, after second-quarter revenue had more than doubled to $11.6 billion.
Signs of the slowdown had appeared before the investor day. Major clients Meta and Microsoft increasingly rely on AI models they build in-house. Anthropic also scaled back distillation, which had brought in 5% to 7% of revenue through paid usage.
Prices are slipping too. Low-cost Chinese labs are adding pressure, and this month Anthropic priced Claude Haiku 5.5 about 75% below its predecessor.
The $2 trillion figure was already doubted by critics well before Wednesday’s numbers. Earlier this month, New Constructs assigned Anthropic a valuation of $150 billion; the research firm also flagged WeWork in 2019.
“We don’t think Anthropic has a viable business.”
Even with a $1.5 trillion valuation, Anthropic would remain under the $1.77 trillion valuation SpaceX reached at its June listing.
The valuation Anthropic lands on also matters outside its own shareholders, because AI sales at Big Tech companies depend heavily on Anthropic and OpenAI.
In August, Steve Eisman estimated that Anthropic and OpenAI account for roughly 70% of AI-related revenue at Microsoft, Amazon, Google, and Oracle. That reliance, he said, was the Achilles’ heel of the AI boom.
Since then, other investors have issued louder bubble warnings. Michael Burry recently said an AI bubble could burst sooner than the 2028 timeline he once predicted.
Arthur Hayes, former BitMEX CEO, likewise expects the AI data center boom to end in a crash and a multi-trillion-dollar bailout.
The news lands during a bruising week for AI stocks. CNBC says the 20 most valuable AI and chip names have lost more than $1 trillion.
Dell fell 6.8% on Thursday, following an 11% drop the previous day. CoreWeave tumbled about 15% on Wednesday.
For the moment, Anthropic stays first in line, while OpenAI is expected to go public by early 2027. Whichever lab lists first is expected to set the industry’s pricing benchmark, analysts say.
Share to
Disclaimer: this article comes from third-party media and is provided for reference only. It does not constitute investment advice. Crypto and other financial products carry significant price volatility risk, so please make your own decisions carefully.
A healthcare technology company is shifting most AI workloads to open-weight and small models, achieving major cost savings while maintaining performance.
Joe Lubin's Consensys sold all its SharpLink shares after a 94% stock collapse. The sale was at $8.39 per share, 93% below the peak.
The Nasdaq slid last week on hot inflation, growth worries, and widening credit spreads. Key focus is Friday's US Flash PMIs.
Germany and NVIDIA signed a joint declaration on industrial AI, with NVIDIA committing nearly $10 billion to European entities.