Wednesday, June 24, 2026

Top story

The NumbersSiliconAngle

Memory maker SK hynix files for $29B US IPO amid AI demand

SK hynix's massive US IPO filing signals that AI infrastructure capex is moving upstream to memory suppliers. HBM (high-bandwidth memory) is now a critical bottleneck in the AI compute stack, and SK hynix—the world's largest HBM supplier—is capitalizing on that scarcity to hit public markets at record valuation.

SK hynix files for $29.4B Nasdaq IPO

The briefs

OpenAI is moving vertically into custom silicon for inference, reducing dependence on NVIDIA and Broadcom while optimizing for its own model workloads. This signals a structural shift: frontier labs are now competing on silicon, not just models.

OpenAI reveals Jalapeño — custom ASIC for AI inference

Google democratizes agentic AI by embedding computer-use capabilities directly into a faster, cheaper model tier. This shifts the competitive landscape away from proprietary agent frameworks toward open, model-native automation—lowering the barrier to entry for enterprises building autonomous workflows.

Computer use capability added to Gemini 3.5 Flash (not just flagship models)

Custom silicon for inference is becoming the competitive moat. OpenAI's Jalapeño signals that foundation model companies are moving upstream into hardware to control costs, latency, and supply chain risk—a shift that reshapes the chip market and threatens NVIDIA's inference dominance.

OpenAI + Broadcom co-developed Jalapeño inference chip

A major semiconductor player is consolidating AI software infrastructure. This signals that chip vendors see compiler/optimization software as mission-critical to competing in inference, and it validates the thesis that AI chip advantage isn't just hardware—it's the full stack.

Qualcomm acquires Modular for ~$4B

China's LineShine demonstrates that domestic semiconductor capabilities can achieve exascale performance, signaling a shift in AI compute infrastructure geopolitics and challenging assumptions about US technological leverage through export restrictions.

LineShine surpasses El Capitan in key performance metrics

A well-funded startup led by a former Tencent AI executive is positioning world models as the next frontier in generative video, attracting serious hardware and automotive players as investors—signaling where capital sees the next wave of AI capability.

Video Rebirth raised $80M in funding

Memory chip demand is now a critical constraint in AI infrastructure buildout. Micron's earnings beat signals that memory, not just compute, is becoming a competitive advantage for companies scaling AI systems.

Micron Q3 EPS: $25.11 (beat Wall Street target of $20.78 by 21%)

Qualcomm is moving beyond mobile into enterprise AI infrastructure with a major validation from Meta, signaling a shift in how hyperscalers source custom compute for training and inference workloads.

Qualcomm unveiled new data center roadmap at Investor Day

Qualcomm is making a strategic pivot from handsets to AI infrastructure through an acquisition and new chip previews, signaling major confidence in its non-handset AI revenue potential and reshaping its competitive position against NVIDIA and custom silicon players.

Qualcomm stock up 14% in after-hours trading