GoKawiil
Tech News
Reports indicate OpenAI and Google are both losing key AI researchers to competitors and startups, though the departures are affecting each company differently in terms of scale and strategic impact. Separately, Nvidia has committed an additional $6 billion to support open-weight AI models, while rare-book dealers report a surge in sales tied to AI-era demand, alongside mixed feelings about that trend.
On his CNBC Investing Club Morning Meeting, Jim Cramer discussed Apple's newly unveiled Mac Mini and Mac Studio, which feature upgraded chips for better AI performance. The new Mac Mini now starts at $899, a $100 increase reflecting higher memory and storage costs passed on to consumers.
Nvidia is gaining backing from more game studios for its RTX Spark technology and ARM-based chip architecture, signaling a shift toward wider hardware compatibility for PC gaming. This move could allow games to run more consistently across different laptop configurations, regardless of the underlying chip design.
Perplexity has released Portable Computer, a version of its agentic 'Computer' platform designed to run entirely on local hardware, starting with Nvidia's DGX Spark desktop and Linux machines with RTX GPUs. The system keeps models, files, and processing on-device by default, only asking permission to route individual steps to cloud-based frontier models when needed, meaning local tasks incur no billing credits.
Nvidia announced that its GeForce Now cloud gaming service will support Valve's Steam Controller and Steam Machine later this year, initially for the Steam Deck and Steam Machine app, with Windows and macOS support coming afterward. The update accompanies other GeForce Now additions, including DLSS 4.5 controls, GOG single sign-on, wider Fire TV Stick availability, and Firefox browser support.
Dick's Sporting Goods shares dropped 15% pre-market after the retailer missed second-quarter earnings expectations and cut its full-year outlook for Foot Locker, whose comparable sales fell 3.6%. Meanwhile, Nvidia announced its Groq 3 LPX rack is entering full production later this year, the first commercial payoff from its $20 billion Groq acquisition, even as its stock posted a seventh straight day of losses ahead of Wednesday's earnings report.
Nvidia reports fiscal second-quarter results Wednesday, with investors closely watching how much of its revenue still comes from a handful of hyperscale cloud giants versus a broader base of AI cloud, industrial and enterprise customers. The company recently began splitting its reporting into hyperscaler revenue and a separate category it calls ACIE, covering thousands of other buyers. CEO Jensen Huang has acknowledged that hyperscalers remain the easiest customers to sell to because there are only a handful of them, compared with roughly 250,000 other companies in the broader market.
Nvidia has released an upgraded version of DLSS Ray Reconstruction, version 4.5, which is now live in 30 supported games. The update refines how ray-traced lighting and denoising are handled, aiming to reduce visual artifacts and improve image stability during gameplay.
Nvidia has increasingly used investments and financing arrangements with AI companies to help drive demand for its chips, effectively acting as a lender and investor to firms that buy its hardware. This financial engineering supports near-term revenue growth but ties Nvidia's fortunes more closely to the success of the AI startups it backs.
Taiwanese prosecutors in Keelung have charged nine people, reportedly including an Nvidia senior manager and two Supermicro employees, with breach of trust and document forgery for allegedly faking records to disguise illegal exports of high-end AI servers to China. Investigators found that of 130 B300 servers falsely documented as installed in Taiwan, 74 were actually shipped to Chinese customers, while 56 were stopped after customs officials spotted inconsistencies. The probe follows the March arrest of Supermicro co-founder Wally Liaw in the US over an alleged $2.5 billion scheme to route restricted AI servers to China.
At Hot Chips 2026, SK hynix's Jaesik Lee said hybrid bonding won't be ready in time for HBM4E and will instead debut with HBM5, as current HBM4 packages are already constrained by a 775-micron thickness ceiling matching standard logic wafers. To keep scaling within that limit, 16-Hi HBM4 stacks now use dies thinned to about 50 microns with half the interlayer gap of 12-Hi designs, while SK hynix continues using its existing MR-MUF bonding method through upcoming Nvidia Rubin-generation products.
Nvidia is working to extend CUDA compatibility beyond x86-64 and aarch64 to RISC-V CPUs, according to a Hot Chips 2026 presentation. The company laid out strict hardware requirements, including RVA23 compliance, adherence to RISC-V server SoC and platform specs, ACPI support, and PCIe coherency, to ensure CUDA runs efficiently rather than falling back on lowest-common-denominator code.