DeepX Bets $100 Million on New AI Chip Using Samsung Technology
THE SO WHAT
DeepX putting $100M behind a 2 nm AI chip on Samsung’s process is another data point that specialized accelerators are not just for hyperscalers anymore. For AI builders, this means more heterogeneity in available silicon—design your inference stack to be portable across GPU and non-GPU accelerators instead of hardwiring to a single vendor.
READ THE SOURCE
MORE FROM THE WIRE
Deep & Emerging TechNvidia Crisis Spurs Scrutiny of Agency That Oversees Chip Curbs
Export controls on AI chips are now a live political battleground, with scrutiny shifting from companies to the agencies setting and enforcing the rules. If your roadmap assumes steady access to advanced accelerators, you need a regulatory risk register alongside your capacity plan.
Deep & Emerging TechBlackstone’s QTS Is Cutting AI Debt Sale by More Than Half
A 60% cut to an AI-linked data center debt sale after investors pushed for higher yields suggests the cheap-capital phase of AI infra buildout is not guaranteed. Infra-heavy AI plays should assume tighter credit conditions and stress-test projects against more expensive, slower debt markets.
Deep & Emerging TechAraceli raises $4.5M for drug discovery tools aimed at reducing animal testing
Funding for cell-imaging tools that replace some animal studies points to a shift toward human-relevant in vitro data earlier in drug discovery. Biotech teams should watch these platforms as a way to de-risk candidates faster and align with tightening ethical and regulatory pressure on animal use.
Deep & Emerging TechUber confirms thousands of driver IDs compromised in data breach
A breach via Fireworks AI’s client databases that exposed thousands of Uber driver IDs shows how AI vendors are becoming new identity attack surfaces. If you’re integrating third-party AI platforms with sensitive data, treat them like core infra vendors—full security review, data minimization, and contractual breach playbooks.