Key Takeaways
- Amazon is reportedly in final-stage discussions to acquire Decart for roughly $7.5 billion, although negotiations could still collapse.
- Decart’s chip-portability software could support Amazon’s effort to expand adoption of Trainium3 and compete more directly with Nvidia.
- The proposed valuation represents a sharp increase from Decart’s $4 billion valuation following a $300 million funding round in May.
Amazon is in advanced discussions to acquire San Francisco-based artificial intelligence startup Decart in a transaction valued at roughly $7.5 billion, according to a Saturday report from the Financial Times. An agreement could be finalized as early as next week, but the talks have not produced a completed deal and could still fall apart.
The potential acquisition would give Amazon control of technology designed to help developers alternate between different AI chips. That capability is becoming more valuable as businesses seek additional computing capacity, manage infrastructure costs and reduce dependence on a single processor architecture.
Decart’s value to Amazon may extend well beyond adding another AI product to its portfolio. Decart was an early adopter of Amazon’s Trainium3 chips, which Amazon launched last year as an alternative to the graphics processing units supplied by Nvidia. Acquiring Decart could give Amazon a software layer that makes its own processors easier to use in mixed-chip environments.
That matters because processor availability is only part of the AI infrastructure contest. Developers also have to move models and workloads among different hardware configurations without extensive reengineering. Decart’s main software product addresses that problem by helping developers switch between chips, according to the report.
If Amazon completes the transaction, Decart could strengthen the broader commercial case for Trainium3. Amazon could potentially integrate Decart’s software more closely with its cloud infrastructure, lower migration barriers for developers and make Trainium3 accessible to customers that still rely heavily on Nvidia hardware. The strategic question is straightforward: can better software portability persuade more organizations to diversify their AI computing stacks?
The reported price also illustrates how quickly Decart’s valuation has risen. Decart was valued at $4 billion in May after completing a $300 million funding round backed by early-stage investors including Sequoia Capital and Nvidia. A transaction at roughly $7.5 billion would place Decart’s value substantially above that May benchmark.
Earlier signals pointed in the same direction, though at a lower price. The Wall Street Journal reported in October that Amazon was discussing an acquisition of Decart worth nearly $7 billion. The latest report suggests that negotiations have progressed and that the prospective valuation has increased by approximately $500 million.
Founded in 2023 by a team of Israeli engineers, Decart is also developing Lucy and Oasis world models. Those systems are intended to support live video editing and the creation of interactive virtual worlds. These projects could give Amazon assets spanning AI infrastructure, developer tooling and generative media rather than chip-management software alone.
Nvidia’s position adds an unusual wrinkle. Nvidia participated in Decart’s $300 million funding round, yet Decart’s adoption of Trainium3 and chip-switching technology could help Amazon compete with Nvidia more effectively. An Amazon acquisition would therefore place an Nvidia-backed startup inside one of Nvidia’s major infrastructure challengers.
Regulatory scrutiny would also be a consideration if the parties sign an agreement. The U.S. Federal Trade Commission evaluates transactions for their potential effects on competition, including acquisitions involving emerging technologies and powerful platform operators. The available reporting does not indicate whether regulators have reviewed the prospective transaction, and no definitive agreement has been announced.
For enterprise technology buyers, the larger issue is whether chip portability becomes easier and more practical. Organizations building AI applications increasingly have reasons to compare hardware based on price, supply, performance and workload requirements. Software that reduces switching friction could give buyers more room to negotiate and more options for deploying models.
Still, the immediate facts remain limited. Amazon and Decart are reportedly approaching the final stages of negotiations, not announcing a completed purchase. Until an agreement is signed, financing, governance, regulatory provisions or valuation could change. If the deal closes near $7.5 billion, however, Amazon would be making a sizable bet that the future of AI infrastructure depends not only on producing competitive chips, but also on making workloads easier to move among them.
⬇️