CoreWeave signs Wayve Blackwell GPU training agreement | Magica
CoreWeave adds Wayve to Blackwell cloud, with price and capacity undisclosed
Editorial Team
••📖6 min read
Wayve is expanding from CoreWeave’s developer software into its Blackwell GPU cloud to train driving models. The multi-year deal adds computing infrastructure alongside an established Azure relationship, but discloses neither capacity nor price.
Wayve will train driving models on CoreWeave’s Nvidia Blackwell GPUs, expanding an existing developer-software relationship.
The deal follows years of work with Microsoft Azure and an automotive production agreement; it does not establish a cloud replacement.
Contract value, GPU capacity and Wayve-specific performance results remain undisclosed.
CoreWeave, the Nasdaq-listed provider of computing infrastructure for AI, announced a multi-year agreement on November 9 to supply its Blackwell GPU cloud to Wayve, a British developer of AI driving software for automakers and fleet operators. Wayve will use it to train foundation models—the underlying AI models for its driving technology—and expand infrastructure for evaluation and simulation, CoreWeave said in its announcement.
The agreement gives CoreWeave an example of selling computing capacity to an existing software customer as Wayve works toward production deployment. Wayve already used Weights & Biases, the AI developer platform CoreWeave acquired in 2025. Neither company has disclosed enough about the new contract to assess its revenue contribution or the cost of the added infrastructure.
Wayve’s Jaguar I-PACE testing in London, pictured in its May 2022 Microsoft collaboration announcement. Archival image. Source: Wayve. Source: Wayve — Microsoft collaboration.
Training a driving system across vehicle platforms
Founded in 2017, Wayve develops AI Driver, software that learns driving behavior from data. Its intended customers include carmakers and fleet owners, with applications ranging from driver assistance to automated driving. The company describes its approach as mapless and hardware-agnostic, designed to work across vehicle platforms.
This is an established development effort rather than a new entrant buying its first training infrastructure. In its 2024 funding announcement, Wayve said it had conducted UK public-road trials since 2018 and partnered with delivery fleets from Asda and Ocado Group.
UC Berkeley and NVIDIA are releasing tools for AI agents to assemble and improve robot programs in simulation. A warehouse operator reports a useful production update, while unsuccessful fixes and benchmark results show why testing, speed and existing infrastructure still matter.
Editorial Team
CoreWeave now says Wayve’s AI Driver learns from real-world data collected in more than 70 countries. That is the stated geographic reach of the data, not the number of countries with deployed driverless services.
The companies describe a development cycle of gathering driving experience, training models, evaluating them and deploying improved versions. CoreWeave argues that long training runs require reliable infrastructure and visibility into model performance. Wayve says access to the cloud will help accelerate commercial deployment. The announcement provides no Wayve-specific measurements of training speed, cost savings or driving safety to establish that effect.
CoreWeave cites MLPerf training and inference benchmarks and inference rankings for other AI models as evidence of its platform’s performance. Those comparisons do not measure the cost or results of Wayve’s driving workload. CoreWeave, also founded in 2017, listed on Nasdaq in March 2025; adding Weights & Biases that year extended its offering beyond computing infrastructure into developer software.
Wayve’s Azure relationship predates the CoreWeave agreement. The disclosed deal does not establish that CoreWeave replaces Azure or becomes an exclusive provider. Source: Wayve — Microsoft collaboration.
Azure remains a material comparison
Wayve’s cloud relationship with Microsoft began in 2020, when it chose Azure. In a May 2022 announcement, Wayve called Azure its primary cloud platform and described joint work on supercomputing for autonomous-driving models. It said its emerging commercial-pilot workloads were reaching the limits of commercially available cloud and edge services.
That earlier announcement anticipated infrastructure capable of handling much larger models and image datasets. It establishes that scaling compute was already central to Wayve’s plans, rather than a need created by the CoreWeave agreement.
A report on the new deal also identifies Wayve as an Azure customer. The disclosed agreement does not establish that CoreWeave replaces Azure or becomes Wayve’s exclusive provider, and it offers no like-for-like price comparison between them.
Physical capacity is another unresolved part of the arrangement. The same report identifies two operational CoreWeave UK data centers launched in January 2025 and a further £1.5 billion planned UK capacity and operations investment announced in September 2025, on top of an earlier £1 billion commitment. Those are broader infrastructure commitments; the report does not allocate them to Wayve or identify where its workloads will run.
Investors also supply the technology
Wayve’s financial backers overlap with its technology suppliers. Its $1.05 billion Series C round in May 2024 was led by SoftBank Group, with Nvidia joining as a new investor and Microsoft participating as an existing investor, Wayve said. Nvidia supplies the Blackwell GPU technology named in the new cloud agreement; Microsoft supplies Azure. Wayve also said SoftBank joined its board as part of the 2024 transaction.
More recently, Wayve announced a $1.2 billion Series D in February 2026 at an $8.6 billion post-money valuation, in a round led by Eclipse, Balderton and SoftBank Vision Fund 2. Microsoft, Nvidia and Uber participated alongside automakers Mercedes-Benz, Nissan and Stellantis. Its broader $1.5 billion funding package included additional Uber capital tied to milestones, rather than solely the closed Series D.
In that announcement, Wayve described its business as licensing AI Driver directly to automakers. It said the driving system runs on onboard computing and sensors. The CoreWeave agreement concerns model development in the cloud, rather than the computing that operates inside a vehicle.
In April, Wayve announced a $60 million extension from AMD, Arm and Qualcomm Ventures. That broadens its silicon-industry backing beyond Nvidia. These financings help explain the resources and industrial relationships behind Wayve’s commercialization effort; none is a disclosed budget for the CoreWeave contract.
CoreWeave’s own funding requirements make contract economics consequential. An accompanying financial report puts its debt at $35 billion and cash at $5.5 billion at the end of June, and cites annual capital-spending guidance of $35 billion to $39 billion. Those company-wide figures describe the scale of its financing needs, not the profitability or funding of this customer agreement.
The same report relays Bloomberg reporting, citing people familiar with the matter, that Microsoft was reassessing external AI computing arrangements amid concerns over costs and delays. Microsoft is also a major CoreWeave customer, making demand from that company material to CoreWeave’s business. The Wayve announcement does not resolve that separate customer-demand question.
Official imagery from Nissan and Wayve’s December 2025 agreement. Nissan targeted the first model with the new ProPILOT generation for Japan in fiscal year 2027. Source: Wayve. Source: Wayve — Nissan agreement.
The next test is production progress and disclosed economics
Wayve has a concrete automotive milestone against which to assess its development effort. In December 2025, it and Japanese automaker Nissan announced definitive agreements to integrate AI Driver into Nissan’s next-generation ProPILOT driver-assistance system. ProPILOT began with single-lane highway assistance in 2016 and added multi-lane support and hands-off functionality in 2019.
Nissan had unveiled a prototype in September 2025 combining Wayve software with its perception technology using next-generation lidar. The companies said Nissan was the first automaker to commit to deploying Wayve’s AI at scale across a broad range of vehicle segments. Nissan targeted the first model with the new ProPILOT generation for Japan in fiscal year 2027. That is a production plan for driver assistance, not evidence of a driverless launch or of acceleration caused by the new cloud agreement.
CoreWeave is scheduled to report third-quarter results on November 10, according to the financial report. For assessing the Wayve deal, the missing disclosures are more specific: committed GPU capacity, service location, rollout schedule, contract value and duration beyond “multi-year.” Measured changes in Wayve’s training and evaluation cycle would then show whether the additional infrastructure improves development economics. Without those details, a broader supplier relationship is established; its financial return and effect on Wayve’s production timetable remain unresolved.
Cloud Pathways lets customers coordinate large TPU workloads from one Python client, extending the system Google used for Gemini training. General availability on GKE follows earlier customer access; substantial reported usage and internal recovery results still leave customer cost savings unmeasured.
NVIDIA’s GR00T 2 predicts video and robot actions together to transfer skills across bodies. Company tests show gains from human video, but selected tasks, corrective data and recovery agents qualify its strongest success rates.
Anthropic says its new AI model produced circuit boards, enclosures and firmware for three prototypes that needed no electrical fixes. The demonstration avoided factory revisions, according to the company, while leaving certification, manufacturing scale and total costs unproven.
Meta is reportedly building a consulting channel for corporate AI sales after announcing an enterprise platform led by former MongoDB CEO Chirantan Desai. Its hiring remit also covers advertising and messaging, while prospective partners already work with rival AI suppliers.
OpenAI’s $200 Chromatic: Codex Edition adds AI-assisted development to an existing cartridge-game platform. The public bundle includes a programmable cartridge, but creation requires a computer, established tools and playtesting; the store now lists a November 19 restock.
PwC US plans a December launch for a critical-thinking and judgment certificate, with completion targeted by June 2027. It will assess staff through performance reviews as the firm seeks new ways to develop skills once practiced through routine work.
Linnea Quigley and Beverly Randolph say Aftermath departed from the musical they agreed to make, exposing a gap between a marketed cast reunion and performers’ control over their screen identities.
Anthropic’s Claude Fable 5.2 and Mythos 5.2 bring lower token prices and stronger technical performance. The release also limits capabilities that could accelerate frontier AI development, as new evidence puts the pace of research close to a sensitive policy boundary.
Meta’s smallest Muse model matches Claude Haiku 5.5’s rounded intelligence score at about one-third of the benchmark task cost. The case for delegating agent work depends on caching, task difficulty and higher charges for long requests.
SpaceX proposes a gas-fired power plant and super-compute facility on about 444 acres in Starbase, Texas. A December hearing will consider local incentives, while the public notice leaves the tax concession and operating demands unspecified.
The California state senator elected to Nancy Pelosi’s open seat wants a federal AI regulator, independent safety checks and penalties for developers. His California record shows both the reach and limits of that agenda; political support from an Anthropic-funded group adds scrutiny.
Apple Creator Studio subscribers can buy extra Logic Pro assistant usage for $49.99 in the U.S. The optional packs last up to a year, but Apple’s public support document does not quantify their capacity.
Boston Dynamics has fitted its electric Atlas humanoid with scooter wheels to test learned controls against real motor limits. The research feeds into the walking robot as Hyundai plans factory deployment, but the demonstration offers no quantified reliability gains.
Nominal raised $240 million at a $3 billion valuation, three times its March valuation, to expand its hardware engineering software into AI agents and lifecycle applications. Its defense foothold and company-reported workflow gains provide a foundation, but do not establish the newer agents’ performance or economics.
The former Megaspeed executive sought backing for a data centre project worth several hundred million dollars, citing GPU access and Alibaba business. The reported pitch adds commercial detail to scrutiny of her ties to Nvidia’s chief, while espionage and chip-smuggling allegations remain unproven.
President Donald Trump offered an unnamed executive’s alleged reversal as evidence that AI risks can be managed. His account identifies neither the conduct at issue nor the legal basis for a jail threat, leaving its value as a safety precedent unproven.
The Texas startup is building factory-made data centers for AI customers, with testing reportedly planned for early 2027. Its financing backs a manufacturing strategy whose cost and power-access advantages remain to be demonstrated.
Baidu says its automated research system has improved inference for the forthcoming ERNIE-5.5-Flash, alongside public benchmark gains. The results show bounded engineering progress, while the cost of finding those improvements remains undisclosed.
