Amazon Weighs $7B Acquisition of AI Startup Decart

Cloud hosting giants are spending fortunes to free their server facilities from single-vendor silicon dominance. For three years, hyperscalers poured hundreds of billions into purchasing graphics processors from merchant suppliers, watching their operating margins shrink as component prices soared. Cloud operators recognized that buying hardware alone could not break vendor dependencies; they needed software compilers that allow developers to run heavy machine learning workloads across custom and merchant chips without rewriting code. Amazon is moving aggressively to lock down that software translation layer. On Sunday, October 11, 2026, The Wall Street Journal reported that Amazon entered advanced negotiations to acquire Israeli software startup Decart in a transaction valued near $7B. The acquisition would represent one of the largest corporate purchases in Amazon history, strengthening Amazon Web Services as cloud rivals battle for computing independence. We examined how massive corporate deals reshape computing infrastructure in our report detailing Nvidia acquiring Hugging Face in a multi-billion dollar transaction.

The deal talks mark an unexpected turn for the San Francisco and Tel Aviv startup. Decart was established in late 2023 by brothers Dean and Orian Leitersdorf alongside Moshe Shalev, who met during military intelligence service in Israel Unit 8200. In May 2026, the company raised $300M in a funding round led by Radical Ventures with equity participation from Nvidia, pricing the venture at roughly $4B. Over recent months, the company became the center of a bidding contest. Nvidia explored an outright purchase earlier in the summer, before American laboratory Anthropic emerged as a preferred bidder in August, discussing a buyout between $6B and $7B. Anthropic walked away in September following due diligence reviews, leaving an opening for Amazon. While negotiations remain fluid and could still collapse, signing a $7B deal would bring one of the premier compiler and simulation teams directly onto Amazon corporate balance sheet. The rapid growth of venture valuations across frontier computing mirrors trends we tracked when Anthropic prepared its public market valuation targets.

Silicon Portability and the Struggle Against the CUDA Wall

To grasp why Amazon is willing to pay $7B for a three-year-old software firm, one must examine the software fortress protecting traditional chipmakers. Nvidia dominates enterprise computing because of CUDA, its proprietary software language that developers have used for twenty years. Programmers write machine learning code tailored to CUDA cores. When a company attempts to run those identical models on alternative silicon, the code fails to compile, requiring months of manual re-engineering.

Decart solved that operational barrier by engineering hardware-agnostic execution engines. The startup software sits between application code and physical silicon, translating mathematical matrices into low-level instructions that execute across Nvidia GPUs, Google Cloud TPUs, and Amazon custom Trainium chips. By eliminating hardware lock-in, Decart gives cloud customers the freedom to shift active workloads to whichever processor cluster is available and cost-effective. For Amazon Web Services, owning this software bridge removes the primary excuse enterprise clients make for refusing to adopt Amazon internal processors. How specialized chip architectures challenge proprietary software platforms was explored in our analysis of Axelera unveiling custom processing silicon for data factories.

Accelerating the AWS Trainium Hardware Pipeline

The acquisition directly supports Amazon multi-year effort to commercialize its proprietary silicon portfolio. AWS has developed multiple generations of its custom Trainium and Inferentia chips, marketing them as lower-cost alternatives to merchant silicon. In 2025, Amazon engineering teams gave Decart early access to Trainium3 silicon, recognizing that the startup compilation software could unlock superior processing speeds from the hardware.

Integrating Decart compiler engineers into the AWS Annapurna Labs hardware team could accelerate the rollout of Trainium clusters across global data regions. If AWS can promise corporate clients that porting existing foundation models to Trainium takes hours rather than quarters, enterprise adoption will climb sharply. Every workload that transitions from expensive merchant graphics cards to custom Amazon silicon widens AWS operating margins, allowing the cloud giant to undercut competitors on hourly compute rental rates. The scale of capital expenditures funneled into server facility infrastructure matches patterns we documented when Crusoe secured $3B to expand physical data centers.

From Video Processing to Real Time Oasis Simulations

Beyond compiler translation layers, Decart established itself as an early pioneer in generative world models and interactive video environments. The company captured international attention by launching Oasis, a foundation model capable of generating interactive, playable digital worlds in real time based on player keyboard and mouse inputs, functioning without a traditional game engine.

The startup also developed Lucy, an advanced neural engine engineered to edit and synthesize high-resolution video streams in real time. World models represent the next frontier in computational research, providing the digital training grounds needed to teach autonomous robotics, drone navigation systems, and automated driving software how physical reality operates. By acquiring Decart, Amazon gains immediate technical capabilities in real-time simulation, providing tools that could support autonomous fulfillment robots inside Amazon warehouses and autonomous vehicle development at its Zoox subsidiary. The intersection of generative models and physical robotics was examined in our report on Skild AI and Nvidia teaching robots through video learning.

The Collapse of the Anthropic Deal and Due Diligence Shadows

The backstory surrounding the failed Anthropic acquisition adds complexity to Amazon negotiations. In August, Anthropic was positioned to acquire Decart for nearly $7B, seeking to secure compiler talent and simulation technology to support its Claude model family. After conducting detailed technical and legal due diligence, Anthropic abruptly ended the talks in September without signing a final purchase agreement.

While neither party publicly disclosed the reasons behind the collapse, industry insiders note that corporate acquisitions of this scale frequently stumble over intellectual property ownership, export control liabilities, or commercial revenue verification. Decart commercial revenue remains modest relative to a $7B price tag, meaning the valuation rests heavily on future strategic utility rather than existing balance sheet cash flow. Amazon corporate development teams must conduct exhaustive audits to ensure that the startup patents and software repositories carry clear legal title before committing shareholder capital to a binding purchase agreement. Managing commercial risk during complex corporate transactions matches dynamics we followed when Manus resumed independent operations after its corporate deal collapsed.

Antitrust Scrutiny and Global Regulatory Realities

Executing a $7B acquisition will test Amazon relationship with antitrust regulators in Washington, London, and Brussels. Regulatory bodies like the Federal Trade Commission and the UK Competition and Markets Authority have monitored Big Tech acquisitions aggressively, investigating whether dominant cloud titans use buyouts to eliminate emerging competition.

Regulators have expressed concern over cloud providers locking up artificial intelligence startups through exclusive cloud credits and equity pacts. If Amazon purchases Decart and restricts its software from optimizing rival clouds like Microsoft Azure or Google Cloud, antitrust watchdogs could launch formal anti-monopoly lawsuits to block the merger. Amazon legal advisors will likely argue that the transaction fosters competition by creating a credible alternative to Nvidia merchant chip monopoly, lowering computing expenses for enterprise software developers across the globe. How regulatory oversight influences sovereign technology deals was explored in our analysis of global powers debating technology development boundaries.

Defending Cloud Margins in an Era of High Infrastructure Spending

The broader commercial motivation behind the acquisition is the defense of enterprise cloud profitability. Over recent quarters, hyperscale cloud operators faced intense scrutiny from Wall Street analysts regarding their soaring capital expenditures on data center land, electrical substations, and high-density server racks. Cloud providers cannot indefinitely spend billions on server hardware without proving that those investments deliver growing software operating cash flows.

Decart software addresses this margin pressure by squeezing greater computational throughput out of existing silicon. If an optimization algorithm allows a cloud cluster to process thirty percent more inference queries per megawatt of electricity, AWS can serve more customers without constructing additional power plants or buying surplus processors. In an industrial environment where electrical grids and server cooling towers are constrained, computational speed directly dictates business profitability. The physical energy bottlenecks facing industrial data centers were examined in our report detailing water and power constraints threatening multi-billion dollar server campuses.

The Race for Sovereign Cloud Platforms

Amazon negotiations with Decart highlight how the technology sector has shifted from developing consumer applications toward building industrial computing platforms. The enterprises that control the software translation tools, the custom silicon fabrication pipelines, and the physical server halls will dictate the terms of global digital commerce for the next generation.

If Amazon successfully closes the $7B acquisition, it will secure an indispensable software asset in its campaign to dethrone proprietary chip ecosystems. By uniting Trainium hardware with Decart compiler software and world simulation engines, AWS can offer enterprise developers a unified, cost-effective computing platform that operates without external hardware gatekeepers. The battle for the future of cloud computing is no longer fought only on price per gigabyte; it is decided by the software intelligence that ties physical silicon together.

Read More on TechRobust:

• Nvidia acquires Hugging Face in $13B deal
• Anthropic IPO filing reveals $2 trillion valuation bet
• Crusoe secures $3B funding round at $30B valuation for data centers