Microsoft opens RTX Spark device preorders and makes Execution Containers generally available
Microsoft opened preorders for the Surface Laptop Ultra and RTX Spark Dev Box and made Execution Containers generally available, letting developers limit agent access to files and networks.
What The Chip Happened To AI & Semiconductor Investors Today? 1) NVIDIA $NVDA / Microsoft $MSFT: Microsoft opened preorders for Surface Laptop Ultra and Surface RTX Spark Dev Box, both built around NVIDIA RTX Spark N1X. The U.S. store lists the entry laptop at $2,599.99 with 24GB unified memory and a 512GB SSD; availability begins October 16. The $5,999.99 Dev Box has 128GB memory and a 2TB SSD, with U.S. shipments planned for November. Alongside the hardware, Microsoft made Execution Containers generally available, letting developers define agent access to files and networks outside the agent’s control. 2) AMD $AMD / Samsung $SSNLF: Lisa Su met Samsung’s semiconductor leadership in Seoul for just over three hours, including dinner, according to Edaily. During the visit, she described tight memory supply, encouraged capacity expansion and discussed planning three to five years ahead. The product connection predates this meeting: Samsung’s March 18 memorandum of understanding (MOU) with AMD outlined primary HBM4 supply for MI455X GPUs and collaboration on advanced DRAM and optimized DDR5 for sixth-generation EPYC “Venice” CPUs and Helios racks, following HBM3E supply for MI350X and MI355X. Foundry cooperation was an opportunity for discussion in that MOU; Edaily’s completed-meeting report disclosed no specific progress toward a foundry deal. 3) Google $GOOGL / Unity $U: Google launched Playground, an experimental browser platform that lets users describe a game, play it and refine it through conversation, then share it by link or through an Explore gallery. It is available to U.S. users aged 18 and over, with creation access rolling out in tiers tied to Google AI subscription. Unity’s Spark integration is forthcoming, with testing underway and a closed beta coming soon. Google describes richer mechanics, 3D capabilities and Unity runtime flexibility; Unity says Spark connects directly to its Asset Store. The intended audience includes people with a game idea but no development experience, alongside artists and modders. That expands the creation approach beyond Unity’s May beta Editor tools for Unity 6+ projects, which helped developers write scripts, change scenes and verify behavior. 4) Wolfspeed $WOLF: The company announced a conditional loan commitment of up to $1.5B from the U.S. Department of War’s Office of Strategic Capital to support domestic silicon carbide materials and wide bandgap power devices. Planned uses include upgrading gallium nitride epitaxy for communications and electronic-warfare systems and developing radiation-hardening capabilities. The contemplated facility is senior secured, allows delayed draws and has a 30-year term. Subject to definitive agreements, Wolfspeed would issue VWAP-based warrants for up to 7.5% of fully diluted equity, pro rata as financing tranches are funded. Funding remains subject to due diligence, agreements, government approvals, appropriations and third-party consents; Wolfspeed says financing is not assured. 5) Samsung HBM4E: Samsung’s 12-high HBM4E reportedly passed qualification at NVIDIA and major hyperscalers in late September, according to Hankyung. Samsung told the publication it could not confirm the report. The prior confirmed step was its May 29 announcement of sample shipments: 48GB capacity and bandwidth of up to 3.6TB/s per stack. Hankyung’s reported milestone is a qualification pass, while customer-specific volumes and mass-production timing remain unconfirmed. Samsung’s July earnings update separately said HBM4 sales had scaled during the second quarter and projected that AI/server-memory demand would keep the market undersupplied in the second half. 6) Aehr Test Systems $AEHR: Aehr’s October 7 announcement reported about $6M in follow-on Sonoma production orders for its lead hyperscale customer’s next-generation AI processor. The order includes systems, burn-in modules, device-specific sockets and software enhancements, with deliveries to a Taiwan outsourced assembly and test partner scheduled over the next six months. Aehr says the customer’s first Sonoma processor is already in high-volume production; its second program is expected to ramp in calendar 2027, with eventual volumes expected to exceed the first. The initial production win for that next-generation processor was announced February 11. Sonoma provides independent thermal control across zones within these high-power packages. For company-wide context, Aehr’s July 14 outlook projected $130–150M of revenue for the fiscal year ending June 25, 2027, versus $50M of revenue in FY2026 ended May 29, 2026. 7) CoreWeave $CRWV: CoreWeave’s October 7 India announcement outlined a planned 240 MW AI cloud deployment with AdaniConneX at Taloja in Navi Mumbai. The campus would span three 80 MW buildings, with CoreWeave as sole tenant, and the company plans to install NVIDIA Vera Rubin. The first phase is expected in mid-2028, followed by additional capacity in stages; a further expansion of up to 240 MW is a separate opportunity. The design includes direct liquid cooling and non-evaporative chilled water. This follows CoreWeave’s August 4 Indonesia expansion announcement: three facilities totaling 360 MW of contracted IT power, expected to become operational in 2028.