Capital

Anthropic Eyes $6B Deal to Cut Its Compute Bill

Bloomberg reports Anthropic is in talks to buy AI-efficiency startup Decart for about $6 billion, a 50 percent premium on its own funding round. The premium, not the price tag, is the real signal.

CapitalBy Servola Tech Desk2026-08-133 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

Anthropic Eyes $6B Deal to Cut Its Compute Bill

Key takeaways

Bloomberg reported on August 13 that Anthropic is in early talks to acquire AI-efficiency startup Decart for about $6 billion, which would be Anthropic's largest acquisition to date; any enterprise leaning on independent efficiency-tooling vendors to control its own AI infrastructure costs should now treat consolidation, not just pricing, as the risk to watch.
The reported price is roughly a 50 percent premium over the near-$4 billion valuation Decart's own May 2026 funding round set only around three months earlier, for a company founded in 2023.
Decart's core product claims to boost AI training and inference cluster efficiency by roughly tenfold; Anthropic is reportedly planning to fold its team into its inference and performance division alongside a broader hiring push for custom-chip engineers.
The talks remain at an early stage and could still collapse, but the premium itself is the signal: frontier labs are choosing to acquire compute-efficiency technology outright rather than simply buy it as a service, ahead of Anthropic's own anticipated IPO.

What Bloomberg Reported on August 13

Anthropic is in talks to acquire Decart AI for approximately $6 billion, Bloomberg reported, which would be the company's largest acquisition to date. The talks are described as being at an early stage and could still fall apart before any agreement is signed, but the timing is notable: it lands just ahead of Anthropic's own anticipated IPO, at a moment when the company is under pressure to show it can control the cost of running its models at scale.

Decart was founded in 2023 by Dean Leitersdorf and Moshe Shalev, two Israeli engineers, and raised a $21 million seed round led by Sequoia Capital in 2024. In May 2026, the company raised $300 million in a round led by Radical Ventures at a valuation near $4 billion.

What Decart Actually Sells

Decart's core product is software that boosts the efficiency of AI training and inference clusters, with the company claiming roughly tenfold gains in usable throughput from the same hardware footprint. That positions it as a layer any buyer of GPU or accelerator capacity could use to extract more compute from the infrastructure it already owns, rather than simply buying more chips.

Its other product is Oasis, a real-time AI-generated video game built on the company's own world-model technology, which shares underlying infrastructure work with its efficiency tools. If the deal closes, Decart's team is reportedly expected to join Anthropic's inference and performance division, alongside a broader Anthropic hiring push for engineers with experience across the hardware and software stack aimed at custom chips.

The Premium Is the Real Story

If a $6 billion deal closes, that is roughly a 50 percent step-up over the valuation Decart's own funding round set only around three months earlier, for a company not yet three years old. Paying that kind of premium for compute-efficiency technology right before an IPO tells you something more specific than 'Anthropic made a big purchase': it tells you compute cost, not model capability, is now the binding constraint even for one of the best-funded labs in the industry, one that presumably has plenty of in-house talent to try to build this itself.

Buying rather than building also removes an independent option from the market. Any enterprise that might eventually have licensed Decart's efficiency layer to optimize its own multi-vendor AI infrastructure would instead find that capability absorbed into a single lab's internal stack, unavailable to competitors.

What This Means If You Buy AI Capacity, Not Just If You Are Anthropic

For any company budgeting serious money for AI inference or training, the read is not 'watch Anthropic' - it is that efficiency tooling is now expensive and scarce enough that frontier labs will pay acquisition premiums to control it outright rather than buy it as a service. Enterprises relying on independent efficiency-layer vendors to keep multi-cloud or multi-model AI costs down should treat consolidation in that vendor category, not just funding rounds, as the risk worth tracking now.

If the deal closes, Claude users may eventually see efficiency gains passed through as pricing or performance improvements. But they will also become more dependent on Anthropic's own infrastructure roadmap for compute economics that, until now, were at least in principle a separately buyable layer.

Frequently asked questions

Has Anthropic actually closed the deal to acquire Decart?

No. As of the August 13 Bloomberg report, the talks are described as being at an early stage and could still fall apart before any agreement is signed.

Why would Anthropic pay more than Decart's own funding round valued the company at?

Decart's May 2026 round, led by Radical Ventures, valued it near $4 billion; a reported $6 billion price would be roughly a 50 percent premium over that figure just months later, reflecting how much frontier AI labs now value proven compute-efficiency technology.

What does Decart make besides efficiency software?

Its other product is Oasis, a real-time AI-generated video game built on the company's own world-model technology, which shares underlying infrastructure work with its efficiency tools.

Does this deal matter if my company does not use Anthropic's Claude models?

Yes, indirectly. It signals that independent AI compute-efficiency vendors are increasingly acquisition targets for frontier labs, which is a consolidation risk for any enterprise depending on a standalone efficiency-tooling vendor to control its own AI infrastructure costs.

The story every outlet led with was the price tag and the IPO timing. The number worth budgeting around is the 50 percent premium Anthropic is reportedly willing to pay for a two-year-old company's compute-efficiency code - a signal that the market for making AI cheaper to run is now valuable enough that frontier labs would rather own it outright than keep buying it as a tool.

AnthropicDecartAI acquisitionscompute efficiencyM&AAI infrastructure

Even a Trillion-Dollar Lab Must Rent Its Compute

Anthropic, valued near a trillion dollars, is in talks to rent about 10 billion dollars of compute from Meta over two years. Why the shortage is structural and what owners should budget for.

4 min read

Capital

Volta Infra: $10bn Anthropic Deal at 7 Months

Volta Infra Holdings, a Nvidia-backed neocloud founded in January 2026, raised $300 million at a $2.4 billion valuation on August 4, 2026, then locked in a six-year, $10 billion compute deal with Anthropic for 133 megawatts at a Bitdeer-operated Norway data center. What a lease-financed GPU access chain means for any owner evaluating an AI vendor or compute commitment.

6 min read

Decisions

The Third Pivot Was the Warning Magic Leap Missed

Magic Leap filed a WARN notice for 193 layoffs effective 1 October 2026, its fourth strategic pivot after $4 billion raised. The lesson is the pattern, not the pivot.