Amazon is reportedly considering a roughly $7 billion acquisition of Decart, an AI startup whose software helps developers move workloads across different chips.
The Seattle giant is in talks to acquire Decart, according to a Wall Street Journal report published Oct. 11. The San Francisco-based startup, founded in 2023 by Israeli engineers Dean Leitersdorf, Orian Leitersdorf, and Moshe Shalev, develops tools that help companies move AI workloads across different processors.
For AWS partners and MSPs, the potential deal raises questions about the tools available to optimize customers’ AI workloads and the effort required to move them between chip architectures.
The talks follow Amazon’s April agreement to acquire satellite operator Globalstar, a separate transaction expected to close in 2027, subject to regulatory approvals and other conditions.
Decart raised $300 million in May at a nearly $4 billion valuation, drawing backing from Sequoia Capital, Benchmark, Radical Ventures, and Nvidia. Beyond chip optimization, Decart develops Oasis, an interactive world model designed for physical AI applications such as robotics, and Lucy, which transforms live video for uses including gaming, advertising, and virtual try-on.
AWS offers its own Trainium processors alongside Nvidia infrastructure. Decart says its Optimization Stack supports Nvidia GPUs, Google TPUs, and Amazon Trainium, helping customers optimize AI training and inference across different hardware.
The companies already work together. In its May 18 funding announcement, Decart said its AWS partnership includes a commercial agreement and joint go-to-market initiative. The startup also said it is a member of the Amazon Partner Network. An acquisition could deepen that relationship. However, the reported talks do not establish whether AWS would introduce new migration services or change customer pricing.
Anthropic previously explored a Decart acquisition
The steep $7 billion price tag carries immediate intrigue. Just last month, AI lab Anthropic, which counts Amazon as a primary investor and cloud backer, abandoned its own $6 billion pursuit of Decart after concluding due diligence, according to TechRepublic.
The reason Anthropic ended the talks remains unclear. For Decart customers and partners, a separate question is whether an Amazon acquisition would change support for rival hardware or access to its optimization tools.
What AWS partners and MSPs should watch
For AWS partners and MSPs advising customers on AI infrastructure, Decart’s technology could broaden the options for optimizing workloads. The business case would still depend on the customer’s model, performance requirements, migration costs, and available computing capacity.
- Measure savings across the full workload. Partners evaluating alternative chips should compare training or inference costs alongside engineering time, software licensing, data transfer, and ongoing support. A lower compute rate alone would not establish that a migration saves money.
- Validate compatibility before committing. Support for multiple chip architectures does not guarantee that every model can move unchanged. A pilot should test model quality, latency, throughput, and integration with the customer’s existing tools before partners recommend a production migration.
- Separate hardware flexibility from capacity guarantees. More processor options could help customers manage dependence on one supplier. Chip portability, however, does not establish automatic switching between cloud providers or guarantee spare capacity when demand rises.
- Watch licensing and partner access. If an acquisition proceeds, partners should look for confirmed changes to Decart’s support for Nvidia and Google hardware, its commercial terms, and its AWS integrations. Those details would help determine whether the technology expands customer choice or creates new dependencies.
For service providers, workload assessment, performance testing, and ongoing cost management are potential areas to explore. The immediate takeaway is to evaluate existing capabilities and base customer proposals on measured results, while watching for a confirmed deal and any subsequent product or partner announcements.
Read more: Learn how AI tokenomics and FinOps can help businesses assess AI spending, measure value, and build stronger cost governance.