AI hardware startup Etched is evaluating investment proposals exceeding $40 billion, with some reaching $50 billion, just months after raising $700 million at a $21 billion valuation. The company is developing proprietary chips aimed at accelerating AI inference, which has attracted significant investor interest due to its potential to challenge Nvidia and its existing customer orders, including from quant trading firm Jane Street. These funding discussions are still in early stages, and terms may change if a deal is finalized.
Walmart has slashed the price of MSI's Codex Z2C gaming desktop to $999, saving $600. The system features an AMD Ryzen 7 8700F CPU, GeForce RTX 5060 Ti graphics card, 16GB DDR5 RAM, and a 1TB PCIe 4.0 SSD, all built on a Pro B840-VC WiFi motherboard with upgrade options. The deal has quickly gained popularity, with over 500 units sold since yesterday.
Reflection has unveiled Beam, its first open-weight model with 501 billion parameters, designed for tasks like coding, reasoning, and agentic workloads. The model was pretrained on 23.8 trillion tokens and refined through extensive reinforcement learning using over 100 million rollouts on NVIDIA GPUs, aiming to achieve high efficiency and performance. Beam is currently undergoing final testing before a public release of its weights and technical documentation later this month.
Former Groq engineers Joshua Rubin and Benjamin Serebrin filed suit on Oct. 2 in Delaware's Court of Chancery, claiming Groq's board agreed to Nvidia's deal without a required stockholder vote and without seeking the best possible price for the company. The deal, announced in December, involved Nvidia licensing Groq's AI inference technology and hiring Groq's CEO Jonathan Ross, president Sunny Madra and other senior leaders, while Groq says it remains an independent company that has since raised about $1 billion. Groq called the lawsuit meritless and said it intends to fight it.
According to the Financial Times, Oracle has agreed to lease about 100,000 advanced AI chips to Tencent across Southeast Asian data centers over five years, in a deal valued near $7 billion with roughly 30% paid upfront. Neither company has confirmed the arrangement, and the specific chip model involved has not been disclosed, though the FT notes such leases are permitted under current US export rules since the hardware is unavailable in China.
At a September gathering of about 50 enterprise customers in Seattle, CEO Satya Nadella urged companies to widely deploy AI tools, describing adoption as happening 'bottom up, top down, and middle up.' He has also recently spoken publicly about pacing AI development responsibly and addressing safety concerns, even as Microsoft hosts a Windows and Surface event in San Francisco focused on 'local AI' for PCs.
Reddit user DallasGrave, reportedly affiliated with cooling company Grave PC, detailed a mod that solders a pickup PCB to the RTX 5090's 12VHPWR header and routes power through an added dual eight-pin board, spreading current across three cable sets. The user says a card modified this way ran at 550W for 48 hours with cable temperatures staying under 40°C, and plans to offer the conversion as a paid service with warranty, with a PNY-branded version also in the works.
Nvidia increased the price of its Shield TV Pro streaming box from $199.99 to $299.99 starting October 2nd, citing rising component costs linked to the ongoing memory shortage. The company has also discontinued the cheaper entry-level Shield TV, leaving the Pro as its only streaming device.
Cerebras Systems' stock gained in premarket trading Monday after OpenAI CEO Sam Altman posted on X calling Cerebras a 'close partner' with 'deep engagement pushing on the frontiers of speed.' The comments follow a 20% drop last week after reports that OpenAI would use Nvidia GPUs rather than Cerebras chips to power GPT-6.1 Sol's 'Ultrafast' mode. Cerebras stock was last up 6.3% at $177 per share, still roughly half its post-IPO high, with market cap near $39 billion versus $95 billion at its May debut.
Microsoft is broadening its Advanced Shader Delivery (ASD) service beyond AMD Radeon GPUs to also support Intel, Nvidia and Qualcomm graphics hardware on Windows PCs, with support arriving later in October. ASD streams precompiled shaders from the cloud rather than requiring each PC to compile them locally, a process Microsoft says it can shrink from several minutes to a few seconds. The feature currently works with Gears of War: E-Day and was confirmed by DirectX program manager Wendy Ho in a company blog post.
For more than twenty years, home consoles relied on dedicated, fixed-function graphics chips rather than general-purpose processors. The NES, for instance, used the Ricoh 2C02 Picture Processing Unit, which could not run code but instead converted cartridge tile art and memory lookup tables into a TV signal line by line, with a hard cap of eight sprites per scanline causing visible flicker in games like Super Mario Bros. This era ended in November 2001 when Microsoft released the original Xbox, built around an Intel Pentium III CPU and an NVIDIA graphics chip derived from the GeForce 3, the first GPU to support programmable pixel and vertex shaders.
Developer SupraGSX has created a custom Nvidia ForceWare 382.69 driver that brings support for Pascal-generation GPUs, including the GTX 10-series and Quadro P2000-P6000 cards, to 32-bit Windows XP. The driver combines code from several older Nvidia releases and adds modern display features like DisplayPort HBR2/HBR3 and HDMI 2.0 support, with most cards supported except the GT 1030 and experimental support for the GTX 1050 Ti.