GoKawiil
Tech News
A truck disguised as a food transport vehicle picked up six pallets of Nvidia hardware worth $11.4 million after they left Oakland bound for Texas on August 29, according to ABC7 News. Investigators lost track of the truck in San Bernardino County, and Oakland Police received a theft report on September 13; no arrests have been announced.
The IEEE Computer Society, a co-sponsor of the CVPR conference, released a report examining whether GPU access and development time influenced which papers were accepted to CVPR 2026. The study was produced by the CVPR 2026 organizing team to look at resource disparities among authors submitting research.
Simplismart, a four-year-old Bengaluru-based startup, raised $14 million in a Series B round led by Dallas Venture Capital, with participation from Nvidia, Accel, Shastra VC, and Micromax. The all-primary round values the company at roughly $100 million and will support its software that helps businesses run AI models more efficiently on existing computing infrastructure.
Elon Musk said on X that SpaceX closed $40 billion in debt financing to acquire GPUs, with delivery expected in the first half of 2027. He made the disclosure while responding to investor Chamath Palihapitiya's warning about financing strain and construction delays leaving AI computing capacity underused next year. Musk said SpaceX secured the funding due to its investment-grade status and revenue from Starlink and launch operations, separate from AI data center income.
Elon Musk announced on X that SpaceX has closed $40 billion in debt financing to buy GPUs for AI computing, with funds expected to arrive in early 2027. He said the financing is mostly callable debt and that SpaceX qualifies as investment grade due to revenue from Starlink and its rocket launch business. Bloomberg had earlier reported the company was negotiating with banks including Goldman Sachs and investors like Apollo Global Management to raise the funds for Nvidia chips.
Business Insider reports, citing internal emails, that Amazon is buying Nvidia GB300 server racks faster than it can power them, spending roughly $5 billion in August alone through its 'Early Rack Receipt' process and about $3 billion monthly since. Some hardware is being stored in intermediary warehouses because AWS sites lack the electricity to install it, and one internal team reportedly said there was no clear timeline for deploying the accumulating GPUs.
Israeli security firm Lava says it identified 32 Nvidia Blackwell Ultra B300 chips operating inside the network of state-owned China Telecom, despite these chips being barred from export to China under U.S. controls. Lava says it detected the chips using Nvidia's own DCGM Exporter monitoring tool, though it cannot confirm exactly where in China Telecom's infrastructure they reside. Nvidia did not dispute the finding but said diversion of controlled chips remains statistically insignificant.
According to SemiAnalysis, Nvidia has denied some neocloud providers access to its GTC Berlin conference after they added AMD GPUs or Google TPUs to their cloud offerings. The analysis firm says this follows earlier claims that Nvidia withheld early chip allocations and financial support from neoclouds that diversified their hardware, while larger hyperscalers have not faced the same treatment.
Nebius chief revenue officer Marc Boroditsky told CNBC that prices per megawatt of AI computing capacity are rising faster than expected, citing a Tuesday capacity auction as evidence. He said Nebius's 2027 capacity would sell out and a similar dynamic could occur in 2028, and argued Nebius isn't exposed to the debt concerns that hit rival CoreWeave. Crusoe CEO Chase Lochmiller separately said he's also seeing rising prices for Nvidia GPUs, including older chips.
A free, open-source application called Shader Compilation Stutter Killer (SCSKiller) precompiles game shaders into the GPU driver's cache before launch, aiming to prevent mid-game freezes caused by on-the-fly shader compilation. It works at the engine level, supporting titles built on Unreal Engine 4 and 5, Unity, RE Engine, and FromSoftware's engines, rather than requiring a per-game whitelist. Testers tried it across several games known for stutter issues to evaluate its effectiveness.
Inception Labs released Mercury 3, a text-diffusion large language model it says outperforms GPT-6 Luna and Claude Haiku 5.5 on various benchmarks while generating over 1,000 tokens per second. The model supports a 1-million-token context window, up from 260K in Mercury 2.5, and is priced at $0.20 per million input tokens and $0.70 per million output tokens, discounted 50% at launch. Early-access partners are reportedly using it for chat support, tutoring, data engineering, and web-browsing agents.
GitHub user Panchovix has published a tool that lets Nvidia RTX 30-, 40-, and 50-series GPU owners modify VBIOS settings without physically flashing the firmware chip. Instead, the configurator stores a modified VBIOS image in the GPU's register key, which the Nvidia driver loads on startup, and it works on both Windows and Linux.