Elissa Murphy

The landmark paper “𝘙𝘦𝘵𝘩𝘪𝘯𝘬𝘪𝘯𝘨 𝘓𝘰𝘢𝘥 𝘎𝘳𝘰𝘸𝘵𝘩” (May 2024, Rewiring America & Evergreen Collaborative) laid it out clearly: "𝘔𝘰𝘥𝘦𝘳𝘯 𝘦𝘯𝘦𝘳𝘨𝘺 𝘶𝘴𝘦 — 𝘦𝘴𝘱𝘦𝘤𝘪𝘢𝘭𝘭𝘺 𝘧𝘳𝘰𝘮 𝘩𝘺𝘱𝘦𝘳𝘴𝘤��𝘭𝘦𝘳𝘴, 𝘌𝘝𝘴 𝘢𝘯𝘥 𝘏𝘝𝘈𝘊 𝘦𝘭𝘦𝘤𝘵𝘳𝘪𝘧𝘪𝘤𝘢𝘵𝘪𝘰𝘯 — 𝘪𝘴 𝘨𝘳𝘰𝘸𝘪𝘯𝘨 𝘧𝘢𝘴𝘵𝘦𝘳 𝘵𝘩𝘢𝘯 𝘶𝘵𝘪𝘭𝘪𝘵𝘪𝘦𝘴 𝘤𝘢𝘯 𝘢𝘥𝘥 𝘤𝘢𝘱𝘢𝘤𝘪𝘵𝘺. 𝘛𝘩𝘦 𝘰𝘯𝘭𝘺 𝘷𝘪𝘢𝘣𝘭𝘦 𝘴𝘰𝘭𝘶𝘵𝘪𝘰𝘯 𝘪𝘴 𝘈𝘐-𝘣𝘢𝘴𝘦𝘥 𝘭𝘰𝘢𝘥 𝘤𝘰𝘯𝘵𝘳𝘰𝘭." But they didn’t stop there. They explained how 𝗹𝗼𝗰𝗮𝗹 𝗰𝗼𝗻𝘁𝗿𝗼𝗹 is the lever that makes AI possible: "𝘛𝘩𝘦 𝘨𝘳𝘪𝘥 𝘮𝘶𝘴𝘵 𝘦𝘷𝘰𝘭𝘷𝘦 𝘵𝘰 𝘣𝘦 𝘮𝘰𝘳𝘦 𝘴𝘰𝘧𝘵𝘸𝘢𝘳𝘦-𝘥𝘦𝘧𝘪𝘯𝘦𝘥. 𝘍𝘭𝘦𝘹𝘪𝘣𝘪𝘭𝘪𝘵𝘺 𝘢𝘵 𝘵𝘩𝘦 𝘦𝘥𝘨𝘦 𝘪𝘴 𝘦𝘴𝘴𝘦𝘯𝘵𝘪𝘢𝘭 𝘵𝘰 𝘬𝘦𝘦𝘱 𝘶𝘱 𝘸𝘪𝘵𝘩 𝘳𝘪𝘴𝘪𝘯𝘨 𝘭𝘰𝘢𝘥𝘴." So yes, AI can manage hyperscalers. Yes, AI can help utilities forecast feeder stress. But to reach its potential, 𝗔𝗜 𝗻𝗲𝗲𝗱𝘀 𝗶𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲 𝗶𝘁 𝗰𝗮𝗻 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗰𝗼𝗺𝗺𝗮𝗻𝗱. 𝗧𝗵𝗶𝘀 𝗜𝘀 𝗪𝗵𝗲𝗿𝗲 𝗗𝗖 𝗠𝗶𝗰𝗿𝗼𝗴𝗿𝗶𝗱𝘀 𝗦𝘁𝗲𝗽 𝗜𝗻 Think of DC microgrids as the 𝗳𝗶𝗻𝗲-𝗴𝗿𝗮𝗶𝗻 𝗱𝗲𝗻𝗼𝗺𝗶𝗻𝗮𝘁𝗼𝗿 in the AI-grid optimization stack. They give AI: ✅ A predictable power layer ✅ A digital twin of local loads, generation, and storage ✅ A clean, harmonics-free, real-time controllable circuit And they allow you to build smarter 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝘄𝗮𝗶𝘁𝗶𝗻𝗴 𝘆𝗲𝗮𝗿𝘀 𝗳𝗼𝗿 𝗴𝗿𝗶𝗱 𝘂𝗽𝗴𝗿𝗮𝗱𝗲𝘀. 𝗖𝗼𝗻𝗻𝗲𝗰𝘁 𝘁𝗵𝗲 𝗗𝗼𝘁𝘀 — 𝗙𝗿𝗼𝗺 𝗚𝗿𝗶𝗱 𝘁𝗼 𝗖𝗼𝗻𝘀𝘂𝗺𝗲𝗿 📈 𝗚𝗿𝗶𝗱-𝗹𝗲𝘃𝗲𝗹 𝗔𝗜: forecasts and orchestrates across districts or cities. 🏭 𝗙𝗮𝗰𝗶𝗹𝗶𝘁𝘆-𝗹𝗲𝘃𝗲𝗹 𝗔𝗜: hyperscalers, EV depots, and smart factories use AI to shape peak loads, pre-cool spaces, buffer with batteries. 🏢 𝗕𝘂𝗶𝗹𝗱𝗶𝗻𝗴-𝗹𝗲𝘃𝗲𝗹 𝗗𝗖 𝗺𝗶𝗰𝗿𝗼𝗴𝗿𝗶𝗱𝘀: this is where everything becomes truly programmable: • Solar and batteries speak DC • EV chargers draw exactly what they need • Cooling, lighting, IT and HVAC are coordinated in software • Peak loads are shaved locally — not pushed upstream DC microgrids 𝗱𝗼𝗻’𝘁 𝗷𝘂𝘀𝘁 𝗽𝗼𝘄𝗲𝗿 𝘁𝗵𝗶𝗻𝗴𝘀. 𝗧𝗵𝗲𝘆 𝗳𝗿𝗲𝗲 𝘂𝗽 𝘁𝗵𝗲 𝗴𝗿𝗶𝗱 𝗯𝘆 𝗻𝗼𝘁 𝗻𝗲𝗲𝗱𝗶𝗻𝗴 𝗶𝘁 𝗱𝘂𝗿𝗶𝗻𝗴 𝗽𝗲𝗮𝗸. 𝗥𝗲𝗮𝗹-𝗪𝗼𝗿𝗹𝗱 𝗧𝗿𝗮𝗻𝘀𝗹𝗮𝘁𝗶𝗼𝗻: 💡 𝘏𝘺𝘱𝘦𝘳𝘴𝘤𝘢𝘭𝘦𝘳 𝘊𝘢𝘴𝘦 DC microgrids feed 380V racks, with battery-backed 1500VDC buses. AI rebalances energy flows in real time. PUE drops from 1.15 to 1.03. No grid upgrade needed — just smarter local wiring. 💡 𝘋𝘦𝘱𝘰𝘵 𝘊𝘢𝘴𝘦 480 buses. 215 MWh/day. Legacy AC cabling: 120 km copper. eLoaded DC grid: 27 km aluminum. AI manages which battery to use, which segment to pull from. Capex drops. Cooling loads shrink. 3 transformer stations saved. 💡 𝘙𝘦𝘵𝘢𝘪𝘭 𝘚𝘪𝘵𝘦 𝘊𝘢𝘴𝘦 DC-coupled rooftop solar, AI-managed BESS, and DC lighting/EVs. ToU peaks shaved by 35%. Grid sees smooth load, not a spike. AI doesn't just optimize bills — it stabilizes the feeder.

11

The IBM/Confluent acquisition is a clear signal that real-time streaming is now foundational to enterprise AI. But it also raises a tough question: what happens when openness and independence become absorbed into a single vendor’s stack? If you're concerned about how this impacts you, I recommend this resource from Cloudera that explains the benefits of an independent, managed, containerized alternative: https://shorturl.at/LYcwD

26

2 Comments

Excited to connect at #GTC2026 next week in San Jose! We have been heads-down building something that the industry is starting to talk about more and more: using sub-THz frequencies to complement, and in many cases fully replace, short-reach optical and copper cables inside the rack. The timing feels right. Not only is the research and academia community catching up to what we have believed for a while now, but industry itself is. Operating in the 0.1-1 THz spectrum, these links can deliver ultra-high bandwidth and nanosecond-scale latency while eliminating the inflexibility and maintenance challenges of physical cables. What we are building delivers: Zero-flop data transfer within the rack. Flexible, reconfigurable topology on the fly. Up to 1 Tbps per link, with sub-50 ns single-hop latency and sub-10 pJ/bit energy efficiency. A path toward truly wireless-defined compute infrastructure The cable mess inside today's AI racks is a real constraint on density, serviceability, and deployment speed. Sub-THz changes that equation. People don't like changes, but it's inevitable. If you are at GTC and working on AI infrastructure, interconnects, or next-gen data center architecture, let's find time to connect. The conversation is just getting started. #GTC2026 #AIInfrastructure #DataCenter #Interconnects #WirelessCompute #AIHardware

19

4 Comments

Microsoft just announced 50 trillion AI tokens served last month ... 5X growth from last year. Bittensor subnets 4, 64, etc incentivized well over 1 trillion served in the last month. 2% of Microsoft. Growing much faster. $TAO is coming for everything. https://lnkd.in/g2PPxA8x

17

3 Comments

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

Today Cerebras announced that Amazon Web Services (AWS) will be deploying Cerebras CS-3s in their data centers. Together, Cerebras and AWS will be delivering the fastest inference solution in the world. It has been an extraordinary 30 days for Cerebras. In February, we announced that we would be powering OpenAI’s fast inference offering. And within days of signing the contract, we delivered the first tranche of capacity to production. And today, we announce a multi-year partnership with the industry’s largest cloud provider, AWS. Together we will be building fast inference solutions comprised of Trainium 3 doing prefill and Cerebras’ Wafer Scale Engine doing decode. Cerebras is now powering the world's #1 AI provider, OpenAI. And will be available in the worlds #1 cloud, AWS. Learn more here: https://lnkd.in/gZhevNtg Now, back to work.

1,648

81 Comments

Shares of Broadcom fell over 5% on Friday after warnings of a dip in margins sparked investor worries that the business might not be as profitable as they thought. After tech stocks saw a stellar run through most of the year, talk of an AI bubble - stemming from circular financing, where companies invest in their customers, and frothy valuations on the promise of strong growth - has made some question the nature of the market. Still, Broadcom's shares have soared past competitors this year on the back of strong orders for its custom AI processors from customers including recently announced Anthropic, which has over $21 billion in order value placed with the chipmaker. For more, read my full story for Reuters below:

19

1 Comment