Anthropic courts Decart as agentic load returns to the CPU
Anthropic advances a stock-heavy Decart deal near 7 billion dollars after founders preferred it over a higher Nvidia bid. Agentic pipelines load sold-out server CPUs, and Horizon's Starry wins a first high-end EV cockpit order.
Artificial Intelligence··Night
Anthropic advances a stock-heavy Decart deal
Anthropic and Israeli startup Decart have exchanged advanced drafts on a deal valuing Decart at about 7 billion dollars, paid mainly in Anthropic shares, with signing expected by early September. Nvidia offered a higher valuation, yet Decart's founders and lead investor Sequoia Capital preferred Anthropic. Decart was founded in September 2023, employs about 100 people and has raised 450 million dollars. Its software is reported to let AI models run up to eight times faster, and the company also builds real-time video generation with robotics uses. It was valued at about 4 billion dollars in a recent round and works on a project basis without an established recurring revenue stream. If the purchase completes it would give Anthropic its first development centre in Israel. Capital is therefore choosing a software speed-up story and an Israel footprint over a higher cash-like bid from a chip vendor, even though Decart still lacks steady recurring revenue.[1]
Agentic pipelines put the server CPU back in front
While that capital chases software acceleration, hardware demand is shifting back toward the general-purpose server processor. AMD vice-president Madhu Rangarajan said seven of the eight stages in a realistic agentic AI pipeline run entirely on the CPU. The demand shows up in supply: Intel has cut client CPU production in favour of server parts and says it is sold out of server CPUs through at least the end of the year, while AMD has doubled its server CPU forecast. Two research papers cited by IEEE Spectrum put numbers on the bottleneck: scheduling changes cut end-to-end latency by up to 1.8 times under sustained load, and adding CPU cores reduced time-to-first-token latency by roughly 1.5 to 7 times. Arm and Qualcomm have announced CPUs aimed at agentic work, and Nvidia has given priority to its Vera CPU. Amazon Web Services has reportedly seen waiting times for CPU server capacity climb as AI workloads press on its cloud. The agentic stack is therefore not only a GPU story; the orchestration stages that sit around the model are consuming scarce server CPUs.[2]
Horizon Starry wins a first high-end EV cockpit order
A third allocation decision sits in the vehicle cabin. Gasgoo reports that Pateo will supply a high-end AI cockpit solution running on Horizon Robotics' Starry chip to an unnamed leading Chinese new-energy vehicle manufacturer. Starry, unveiled in April, is presented as the first Chinese-made chip to combine cockpit and driving functions on a single die. Gasgoo describes the order as the industry's first designation for a high-end AI cockpit powered entirely by a domestic cabin-driving integrated chip. On the claims the company supplies, the chip cuts in-vehicle hardware footprint by roughly 50 per cent, lowers total hardware cost by 1,500 to 4,000 yuan per vehicle and shortens the development cycle from 18 months to 8. These figures rest on the maker's own statement, with no independent measurement available, and no mass-production date was given. Together with the Decart talks and the server-CPU squeeze, the day shows capital and design wins chasing different silicon roles: software that claims multi-fold model speed-up, scarce CPUs for agentic pipeline stages, and a domestic cockpit-driving die for a high-end electric vehicle cabin. Throughput marketing is not the shared thread; who gets capacity, a development centre or a vehicle designation is.[3], [1], [2]
