Tony Tzeng

Amazon is rolling out Trainium3 chips in 2026 which can deliver 4.4x more compute and 4x greater energy efficiency than Trainium 2. We expect customer adoption to expand beyond Anthropic in 2026, possibly providing the Street with more proof points that Amazon has a compelling hardware offing. Also press reports have suggested that Amazon and OpenAI on in discussions for a potential deal to use Trainium chips to run AI workloads.

1

Snowcap Compute Inc. is rethinking compute architecture from the ground up and tackling one of the hardest challenges in AI infrastructure: How to dramatically improve performance and efficiency as workloads scale. By advancing superconducting technology from the lab into real-world systems, they’re pursuing a true step-function innovation in how compute is built. Through Playground Global, I’ve had the privilege of backing teams redesigning foundational systems. Snowcap reflects that mindset: disciplined hardware engineering, deep technical rigor and the experience to challenge long-held assumptions in computing. The AI and Quantum eras will demand breakthroughs across silicon, systems, software, and energy. Increasingly, that future will be a cryogenic temperatures. Snowcap is making it really cool to be really cold! https://lnkd.in/gu35jNh8

82

7 Comments

AI is not a singular technology, nor is the compute that powers it. Silicon diversity is not only key for supply chain resiliency but also efficiency. To be profitable, AI must be delivered at scale for the least amount of energy consumed. Large-scale GPU clusters are not always the best answer. CPUs have always been the workhorses of enterprise IT, and now they are also the engines of hyper-efficient AI inference. AMD CPUs are not a commodity. They are the global computing gold standard, ranking #1 in 19 out of 20 categories on Amazon. Read more on TechSpot: https://lnkd.in/gEJ__CWj #Ryzen Dan McNamara Priya Vasudevan Curt Waltman Robert Hoffman Sam Naffziger Andrej Zdravkovic Kumaran Siva Suresh Andani

795

13 Comments

Thrilled to be partnering with Amazon Web Services (AWS) to bring Cerebras speed to all AWS developers. Watch David Brown's video to learn more about our combined solution and get ready to build faster ;)

70

The Brookings Institution has identified five distinct AI clusters in US metros: 🌟 Superstars 🌟 Star Hubs 🌟 Emerging Centers 🌟 Focused Movers 🌟 Nascent Adopters Standing out as an AI Star Hub, the Phoenix-Mesa-Chandler metro area ranks 17th nationally. 🚀 Anticipating positive spillover effects from TSMC's expansion in Arizona, benefiting both the semiconductor ecosystem and AI sector. #AI #Innovation #aidatacenters #semiconductors State of Arizona City of Chandler, Arizona City of Phoenix City of Mesa City of Tempe

10

Quantinuum’s planned IPO is the moment quantum finally hits the public markets in grown‑up form. This isn’t another early‑stage SPAC story; it’s a full‑stack platform spun out of Honeywell, with trapped‑ion hardware that regularly tops industry benchmarks and a real software and applications business on top. By combining industrial‑grade engineering with production‑grade tooling in chemistry, security, and optimization, Quantinuum gives investors exposure not just to “qubits,” but to an integrated quantum stack that already serves enterprise workloads. For technical and business leaders, as well as investors, the key question now is how public markets will price a quantum company that actually looks like an infrastructure and software provider rather than a speculative science project. Whatever the final valuation, this IPO will likely become the benchmark for what “mature” looks like in quantum—deep IP, differentiated hardware, and diversified revenue experiments before ringing the bell. Headline: Honeywell’s Quantinuum Quantum-Computing Unit Plans IPO – https://lnkd.in/euZjkGtG

194

5 Comments

I published a new piece with Silicon Foundry on a pattern that keeps repeating in AI infrastructure. Market leaders rarely miss these shifts because they’re asleep at the wheel. They miss them because they’re doing what leaders are trained to do. They run the playbook that made them leaders. They optimize the system they already have. And they trust the signals that have always correlated with strength. The problem is that AI infrastructure is not just more of the old stack. When architectures move, the basis of competition moves with them. Categories get redefined quietly, then all at once. By the time the change shows up cleanly in the numbers, the advantage has already migrated to a different layer of the value chain. In the article, I walk through why this happens and what to do about it. The ability to notice when the stack is reorganizing, before your planning process forces the answer back into last year’s categories. If you’re operating in semis, networking, or data center infrastructure, I’d be interested in how you’re watching for these shifts in your own system. https://lnkd.in/gimKJPUu

16

3 Comments

GTC 25, Washinton D.C.: NVIDIA CEO Jensen Huang underscored NVIDIA's partnership with Palantir saying Palantir Ontology is the "single most important enterprise stack in the world today." The collaboration is focused on government, national security, and global enterprises. Accelerating data processing "at a much larger scale and speed" is critical for time-sensitive use cases like: * Real-time predictive maintenance for critical infrastructure * Rapid supply chain disruption response * High-volume fraud detection and anomaly mapping * Operational intelligence and decision-making in defense https://lnkd.in/g9rTR23e

7

AI infrastructure is evolving quickly, and NVIDIA is making one thing clear: shared KV cache is becoming essential for the future of inference. But the real challenge is figuring out how to get there from today’s GPU environments without slowing innovation or starting from scratch. Betsy Chernoff shares how WEKA helps teams take a practical path forward, delivering performance gains now while building toward AI Factory and ICMS architectures. If you’re thinking about where AI infrastructure is headed, this is worth a read: http://spr.ly/6045hYo1U

10

Bain: Skyrocketing AI demand creates unattainable CapEx requirement at current compute cost; major technological innovation is needed to reduce cost of compute to keep up. (Spoiler: Investing in some of this at Jumpspeed Ventures ...) https://lnkd.in/dcS8Xddy

12

Building on top of large language models is fun, but getting consistent performance and reliability is extremely challenging. I love this post from Kimberly Patron about how Sierra uses health and performance-based traffic routing and request hedging to get meaningfully better performance and tail latency from high scale LLMs https://lnkd.in/g4qWRCb5

284

13 Comments

🤖 From Robotaxis to Physical AI Autonomy gets attention, but the economics determine what lasts. I built a lightweight Robotaxi Cost Model to explore how utilization, deadhead, and operator-to-vehicle ratios drive cost per mile: 👉 https://lnkd.in/gYa2biDn Big takeaway: once safety is solved, utilization and ops efficiency move margins more than incremental autonomy or hardware cost gains. The same patterns show up across Physical AI — mobility, logistics, infrastructure. Always interested in how others are approaching ops and utilization in Physical AI. #Robotics #PhysicalAI #AutonomousVehicles #FleetOperations #UnitEconomics

15

3 Comments

Jensen Huang says compute equals revenues. This week's filings say compute equals commitments. NVIDIA reported record revenue and a $95.2 billion purchase obligation stack: 5.9x last quarter's revenue locked into forward contracts. CoreWeave beat on revenue and missed on everything else: wider losses, margin compression, Q1 guidance 15% below consensus. Both stocks sold off on strong numbers. The market has stopped asking "is demand real?" and started asking "what happens when the commitments come due?" In this new post, I examine what this week's filings reveal about the structural risk I've been mapping over the past weeks. The bull case requires several consecutive quarters of accelerating revenue. The bear case requires one miss. Did CoreWeave just have that miss? Substack post: "Compute Equals Commitments" - Link in comments.

9

2 Comments

In the latest episode of the OpenAI podcast I got talk to OpenAI CFO Sarah Friar and legend himself Vinod Khosla about the AI industry. Sarah explained the relation between compute and revenue (OpenAI's revenue keeps 3x'ing each year) and Vinod provided an insightful way to tell if we're in a bubble: API calls. Are they going up or down? (Spoiler: Up, up, up!) https://lnkd.in/g2tTQBdp

24

1 Comment