Research / Topics
GPUs, XPUs, CPUs
Merchant GPUs, custom accelerators and the CPUs that sit beside them. Nvidia, AMD, Broadcom, Marvell and the AI chip startups.
61 posts
No posts match.
2026
- AI With No Decode (Jev), and What It Means for AI Infrastructure
Jev from TypeSafe answers in one pass instead of writing text token by token. Why LLMs are hard to use inside software, how Jev fixes it, and what a model with no decode phase means for AI Infra
- 🎙️The Race to Build the Next Trillion-Dollar AI Chip Company
TechSurge: Deep Tech VC Podcast
- A Billion Agents Need Somewhere to Run
Grok Bot, Meta Muse, and what consumer adoption means for CPUs, accelerators, memory, and advanced packaging.
- Open Models & American Diffusion; Nvidia & Poolside
America needs agentic AI to diffuse. Diffusion needs competitive frontier open models.
- Why Did AMD Buy Taalas?
Does it move AMD stock? Speculation. Helios plus Taalas cluster, what about long-context, who uses it, and why buying a one-model chip is less crazy than it sounds.
- The Next Trillion-Dollar Chip Company
It might be simpler than you think. Etched, Fractile, MatX, Positron, Rebellions, SambaNova, Tensordyne, Tenstorrent, and more.
- The ISA Doesn't Matter Where It Counts
x86's lock-in is real. It's just furthest from the GPU.
- Are Agentic CPUs a Commodity? It’s Complicated.
Five sockets, different economics. A model for telling them apart, and where each competitor fits. AMD, Intel, Nvidia, Arm, Qualcomm.
- An Interview with the Gimlet Labs Team About Heterogeneous Inference for AI Agents
Why most neoclouds can't follow Gimlet's silicon-vendor-neutral model, d-Matrix Corsair + NVIDIA B200 delivering 4× Pareto frontier shifts on GPT-OSS 120B, and more
- An Interview with MatX CEO Reiner Pope About LLM Chips
Hybrid SRAM + HBM, MoE interconnect, why frontier labs consider AI ASIC startups, and more
- Agentic AI Needs CPUs. Whose CPUs?
Nvidia Vera, Arm AGI CPU, Meta, x86, more
- The Multi-Silicon Era Is Here
Disagg is out of the bag. What it means for Nvidia, CPUs, XPUs, startups, and more.
- GTC 2026 Keynote Debrief
Scale-up copper & optical, Groq, CPUs, tiered inference economics, SaaS is not dead
- Meta's MTIA Roadmap
Inference-centric design tradeoffs are smart. ROIC is clear. Implications across the supply chain.
- The MI450 Waiting Game
AMD delivered its best quarter ever. Yet the market wants to know why to own it right now.
- An Interview with Microsoft's Saurabh Dighe About Maia 200
The why behind Maia 200. SRAM–HBM balance, scale-up without scale-out, 30% lower TCO claims, and more.
- Right Systems for Agentic Workloads
Why Etched and MatX may be better positioned than Cerebras and Groq for long-context coding agents.
2025
- Right Tool for the Job
Adding a hyperspeed SKU to Nvidia's portfolio
- Nvidia, AMD, Amkor, Arista @ UBS Tech Conference
XPU vs GPU longevity and replacement cycles. Arista's AI opportunities & Broadcom. Amkor Arizona & TSMC.
- Nvidia Q3 Earnings
Great print, Trailing indicator of AI bubble, Value capture, AI lab profitability concerns, Networking business, Meta, China & Taiwan, Google, and more
- AMD & Apple Earnings
What to make of AMD until MI450 is here? And a vintage Apple quarter, but what about AI?
- When Broadcom’s Customers Start Buying Chip Startups
Will these acquisitions disintermediate Broadcom?
- Right-Sized AI Infrastructure. Marvell XPUs
Diversity in AI workloads demands diverse AI datacenters. Sometimes, custom silicon is the answer.
- Nvidia's Climb to $4T and the Peaks Ahead
Inference is scaling, autonomy and robotics are trillion-dollar markets, strength of annual cadence + limited supply, China's big future upside & risks
- AMD's Strategic Bets
FP6 Advantage, Rack Density, HBM, AI NIC, TCO, Openness, UALink, more. Customer Momentum.
- Enterprise AI: Nvidia, Dell, and Implications for AMD
Nvidia vision, Dell validation, AMD questions.
- Dispelling DeepSeek Myths, Studying V3
Training Cost and PTX misunderstandings. Mixed precision FP8 training, communication optimization wizardry, and more
- Nvidia 4Q25 Earnings
Inference-First Deployments, Refresh Cycles, Ethernet, GPUs vs ASICs, and more
- DeepSeek Mixture of Experts & V2
Rewinding: DeepSeek’s 2024 research and its impact on the semiconductor industry
- AMD's AI Accelerator Business: Puts and Takes
Bears, bulls, and timing. It’s all about the long-term view.
- Reasoning Models and DeepSeek, Semis Implications
Reasoning Models, Compute Demand, Scaling Laws, Competition
- CPU Server Growth, Enterprise AI Agent Adoption Challenges
A Case for CPU Server Growth, Enterprise AI Agent Usability, Business Model and Process Challenges
- The Power of Nvidia's Systems Thinking Approach
Systems Thinking, AI Engineers as the Primary Customer, Vertical Integration, Beating Competition
2024
- AMD's AI Woes & Trust Crisis
Concerns and Next Steps After the SemiAnalysis Report
- MatX Prediction
MatX surpasses $100M in funding.
- X86 Commoditization and Competition
Desperate times call for desperate measures
- Groq's Business Model, Part 3: Competing with Google, AWS, Microsoft
Advantages and Disadvantages of Competing Against The Big Dogs
- Groq's Business Model Part 2
Managed Services, Colocation, Traditional Hardware Sales.
- Groq’s Business Model, Part 1: Inference API
Simplifying The User Journey, Business Model Innovation, Customer Benefits
- Chipstrat Chat with Groq's Mark Heaps
API business and "ramp to hardware" model, differentiation, next-gen chip, and more
- ZML's Hardware-Agnostic Distributed Inference
Contemplating the trade-offs and implications
- Semis Funding Heats Up, Groq, Block and China
Money is flowing into logic chips, Groq's Series D, Chinese sanctions play into Block's 3nm ASIC strategy
- AMD's Earnings: Playbook Strength and New Strategic Update
Strategy update, Microsoft wins, Llama 3.1 assist, Demand concerns, Training worries, Client flex, Semi-custom drag
- Jack Dorsey's Bitcoin Push: A Noble Effort, But Will It Matter?
Block’s Bitcoin Mining Rig, Self-Custody Wallet, Aspirations, and Bitcoin Adoption Challenges
- Hardware as a Differentiator: Block's Unique Approach to Financial Services
Hardware as a Core Competency, The "Apple Ethos" in Payments, Business Model, Ecosystems and Playbooks
- Arm's Journey to Value-Based Pricing
Overcoming pricing strategy missteps
- Etched: Silicon Valley's Speedrun
A Novel Semiconductor Startup Fundraising Framework, Etched's Unconventional Approach, and a MatX Prediction
- Intel CPU follow-up, Nvidia Superchips, AMD MI300A
Intel Inside (AI data centers), Grace Hopper & Grace Blackwell, MI300A
- Is Intel's CPU Role Diminishing In The Age of AI?
Is Intel's Datacenter AI Tailwind Slowing? What Pat G Says.
- Lisa Su - Steady As She Goes
Lisa Su's Recent Stratechery Interview And Computex Keynote
- The Rise of Groq: Slow, then Fast
Distractions, Business Model Pivot, Return On Luck, and Looking Ahead
- Rethinking AMD
Modular vs integrated, "right tool for the job" value prop, rethinking the org chart
- Can AMD Bridge Nvidia’s Software Moat?
Transient Moats, Betting on Generative AI, AMD's Path Forward
- 🎧 Chipstrat + The Circuit Podcast
AI Chips: Adoption Challenges, Startup Opportunities, Groq, Hyperscaler Custom Silicon, AWS' Strategy, Legacy Architecture Baggage, and more
- Nvidia and Microsoft - The Impact of LLM-Powered Copilots on Microsoft's Hardware Strategy
Running Microsoft 365 Copilot and GitHub Copilot on Nvidia GPUs ain't cheap. CapEx and OpEx will continue to rise, compelling Microsoft to explore a switch from GPUs to AI ASICs.
- Amazon's Nvidia Entanglement
AWS will continue buying Nvidia GPUs even as competition improves. AMD and AI ASICs like Groq can compete for a spot in AWS's portfolio.
- Nvidia’s Risk
Nvidia's overly concentrated revenue is at risk. Could a custom ASIC business help?
- GPU Bloat Stifles AI
Why Jarvis won’t run on GPUs, and how AI silicon startups like Etched can help
- ASICs: Custom Chips to Improve Performance
The benefits of ASICs. How custom silicon can make a company's "beer taste better". AI accelerators. The rationale behind Broadcom and Marvell's ASIC design service.
