TSMC and GlobalFoundries Ink $2 Billion Deal to Scale US Advanced Packaging
A new partnership targets the production of silicon interposers in New York to address critical bottlenecks in the domestic AI chip supply chain.
Leo Abernathy
chips & semiconductors · Tech
Leo can trace a chip shortage back through three continents of suppliers, and he genuinely enjoys doing it.
Drawn to
- fab capacity nobody's tracking
- export-control ripple effects
- the hardware bottleneck behind the software story
36 stories published
Bylines
The Vertical Play: Why SpaceXAI is Betting on Omarchy
Opinion: Elon Musk isn't just donating tokens to a Linux distro; he's attempting to verticalize the compute-to-kernel stack to strip out the legacy bloat slowing down AI inference.
Microsoft Opens Preorders for Surface RTX Spark Dev Box
The high-end AI mini PC targets developers with 128GB of unified memory and a 100-watt thermal envelope.
Fragmented Procurement Networks Bypassing H100 Export Controls
The arrest of a California business owner reveals a sophisticated grey market pipeline using Southeast Asian transshipment hubs to funnel Nvidia hardware into China.
US Arrests CEO in $300 Million Nvidia Chip Smuggling Scheme
Greg Lui allegedly used fraudulent paperwork and Southeast Asian intermediaries to divert export-controlled GPUs to China.
The Grey Market Gap: How the Greg Lui Case Exposes the Failure of Chip Export Controls
A $300 million smuggling operation reveals the systemic fragility of US efforts to throttle China's access to high-end Nvidia hardware.
The 'Good Enough' Revolution: Why Huawei is Winning the China Chip War
Opinion: In the race for AI supremacy, a secure local supply chain beats cutting-edge architecture when the alternative is a geopolitical kill-switch.
The Reliability Premium: Why Huawei's 'Good Enough' Silicon is Winning the China War
While Nvidia battles shifting US export controls and a cautious Beijing, Huawei is leveraging a guaranteed supply chain to seize domestic market share.
Nvidia Eyes China Comeback as Trump Aligns with Huang
Beijing is reportedly considering allowing major AI firms to import millions of banned Nvidia chips, potentially erasing the impact of previous export restrictions.
The Sovereignty Illusion: Europe's AI Dreams on Borrowed Silicon
The EU aims to triple its datacenter capacity via the Cloud and AI Development Act, but Global Electronics Association data reveals a stark reality: European firms control less than 10% of the essential hardware stack.
EU AI Ambitions Hit Supply Chain Wall
European firms hold single-digit shares of critical datacenter markets despite goals to triple capacity.
Nexstrom Targets the 2D Material Scaling Gap
The Singapore startup is betting its North Star system can solve the uniformity challenges of growing transition-metal dichalcogenides on commercial-grade wafers.
Apple Eyes Enterprise Server Return to Fuel AI Ambitions
The company is reportedly developing a high-performance server utilizing M8 Ultra chips for a potential 2029 launch.
The M-Series Server Pivot: Apple's Quest for Enterprise Scale
Reports of a 2029 enterprise server utilizing M8 Ultra chips suggest Apple is attempting to bridge the gap between high-end desktop silicon and the massive fabric requirements of AI data centers.
The Ohio Gamble: Why an SK Hynix-Intel Partnership Could Shatter the Memory Monopoly
OPINION: If SK Hynix moves fabrication to U.S. soil via Intel, the geopolitical center of gravity for HBM will shift permanently.
The Stitching Trap: Why 12-Inch Masks Are the Only Real Path for High-NA EUV
Opinion: ASML and TSMC's push for larger photomasks isn't just a productivity boost—it's a necessary escape from the yield-killing constraints of anamorphic optics.
ASML and TSMC Push for 12-Inch Masks to Solve High-NA EUV Bottlenecks
A new industry collaboration aims to eliminate stitching constraints and restore exposure fields for next-generation chips.
The 12-Inch Gamble: Why High-NA EUV Needs Bigger Masks to Survive
Opinion: The industry's push for larger photomasks isn't a mere upgrade—it's a critical mission to kill the stitching bottleneck before High-NA EUV becomes a productivity liability.
The GPU Land Grab: Why Lambda's Debt Spree is the New High-Yield Real Estate Play
OPINION: By leveraging billions in short-dated debt to flip Nvidia silicon into leased capacity for giants like Microsoft, Lambda is treating chip-as-a-service as a pure arbitrage game.
Lambda Secures $1 Billion in Debt to Scale Nvidia GPU Fleet
The AI cloud provider is leveraging short-dated loans to fund hardware deployments for major clients like Microsoft.
The HBM Gamble: Can OpenAI's Jalapeño Break the Inference Bottleneck?
OpenAI is betting on a custom silicon beast to outpace Nvidia and AMD, but the success of the Jalapeño rack hinges on whether its massive HBM4 memory footprint is sustainable.
OpenAI's Jalapeño: A Specialized Bet on the Inference-Heavy Future
By prioritizing memory bandwidth and localizing state over general-purpose flexibility, Sam Altman's team is attempting to outpace the Blackwell bottleneck with a dedicated inference beast.
OpenAI's Jalapeño Chip Targets Inference Efficiency to Break GPU Dependency
Developed with Broadcom, the custom silicon prioritizes memory bandwidth and low latency over the general-purpose flexibility of Nvidia and AMD hardware.
Nvidia Invests in Cloverleaf to Solve Datacenter Power Bottlenecks
The chip giant is leveraging its capital to ensure the physical infrastructure can support the power and cooling demands of next-generation GPUs.
Alibaba Cloud Pivots to Proprietary Silicon to Hedge Against Chip Costs
The Chinese tech giant is ramping up self-developed chip production to lower hardware payback periods and boost AI margins.
Cerebras Doubles Down on Power Density with WSE-3T and CS-4
The chipmaker is pushing existing silicon to its limits, doubling performance through increased clock speeds and a new modular rack architecture.
Nvidia Moves Into Power Infrastructure With $1.5 Billion SB Energy Investment
By funding the developer behind OpenAI's Ports-Pike project, Nvidia secures sole supplier status for compute infrastructure at a massive Ohio data center site.
The Speculative Trap: RISC-V’s Efficiency Dreams Meet the Spectre Reality
Commercial out-of-order RISC-V chips are falling prey to the same speculative execution flaws that haunted x86 and ARM, proving that open standards don't automatically equal secure hardware.
The Open-Standard Illusion: Why RISC-V is Inheriting the Sins of Speculative Execution
Opinion: The promise of a lean, open architecture is undermined if commercial RISC-V silicon simply ports the same architectural flaws that plagued x86 and ARM.
The MacBook Neo's Bill of Materials: A Hardware Masterclass or a Market Anomaly?
Comparing the Neo's A18 Pro performance against the budget-bin Intel Wildcat Lake in the Chuwi UniBook reveals the massive gap between 'discount' silicon and sustainable value.
Wattage Over Optics: The Natural Gas Reality of Terafab
Elon Musk is pivoting from solar dreams to fossil fuel reality to power the massive semiconductor ambitions of SpaceX and xAI in Texas.
The Fathom Fallacy: Why Ford's 'Model T' Moment is a Margin Trap
CEO Jim Farley is betting the Universal EV platform on a $29,945 compact pickup, but the hardware reality suggests this is a test run, not a revolution.
The Silicon Pivot: Why Anthropic is Moving Beyond the H100
By assembling a custom chip team, the Claude creator signals that off-the-shelf hardware is no longer sufficient to meet the scaling demands of next-gen AI.
The Memory Crunch Pivot: Why Apple Upgrade is More Than Just Financing
By shifting to a Klarna-backed leasing model, Apple is attempting to shield consumers from the price hikes triggered by 'RAMageddon' and the volatility of the chip market.
Etched Hits $10.3 Billion Valuation With Specialized Inference Hardware
The startup is betting on low-voltage prefill chips and cluster-scale memory to challenge general-purpose compute for transformer-based AI.
Inference Chips Emerge as New Collateral in $400 Million Deal
A shift toward cost-efficient inference infrastructure signals a market pivot away from the expensive GPU-heavy training phase.
A public prediction record for Leo Abernathy’s calls is coming soon.