The AI bottleneck isn't GPUs.
It's power that's already on — EV charging sites, flexible grid load, and home data centers.
Chip fundraising is loud right now, and mostly un-investable from here. Big checks. No lever I can actually pull.
The moat isn't the GPU. It's the energization date.
YC Liquid Energy launched NanoDC: factory-built solar+battery AI boxes that skip the grid queue. Deploy in ~4 weeks, on in 30 min. Interconnect is the new moat — startups are tunneling under it. Does this scale past demo?
teapawt.net/_LiquidEnergy/status/2…
We built a new type of datacenter for AI inference. We call it NanoDC.
NanoDC is a solar-powered autonomous datacenter that can be deployed anywhere on the planet in 4 weeks.
11 Trillion will be invested in AI compute over the next 4 years. Most of this will be for inference and most of it will be severely delayed: stuck in grid interconnect queues, site permits, or supply chain issues. NanoDC sidesteps these issues completely.
NanoDC is built rapidly at a factory and shipped to the site ready to be turned on in 30 minutes. It generates its own power with solar and batteries. You open the door and the solar panels fold-out like an accordion. The compute is cooled with patent-pending high-density air cooling. It connects to the internet via Starlink.
No construction. No grid interconnect. No Water supply.
It never waits for transformers and turbine generators either.
Getting more compute is as easy as dropping more units and turning them on.
The first units will be live this month. Thousands will be shipping starting 2027.
Sunlight in. Tokens out.
Nordic put Wi-Fi, BLE, Thread + an edge AI NPU on one chip — up to 20× lower standby vs prior Wi-Fi parts. Battery devices can finally think locally without a power brick. Next bottleneck: models tiny enough to ship?
nordicsemi.com/Nordic-news/2…
Mercedes just handed its 47 European fast-charging hubs (374 plugs) to Electra. Carmakers love announcing charging networks. Running them is a different sport. My bet: scale and uptime win, so more OEMs quietly hand off. Who's next?
electrive.com/2026/11/03/ele…
A 3-month-old chip startup with no finished chip is in talks at a $500M valuation. Tenspresso's ex-Google TPU team targets long-context inference. Counterpoint to 'bubble': in AI silicon, the team is the product until tapeout. Fair?
bloomberg.com/news/articles/…
Italy's grid operator offered up to $41.8k per MWh-year for battery storage. Winners averaged about $14.6k. Takeaway: cells aren't the bottleneck, cheap money is. 15 years of fixed payments and developers undercut the cap. US utilities, copy this?
teapawt.net/srndptme/status/211804…
Italy offered up to $41,800 a year for battery storage. Developers bid $14,600.
In September 2025 Terna, Italy’s grid operator, asked for 10 GWh of battery storage in the south and on the islands. Terna offered to rent that capacity for 15 years, at up to $41,800 per MWh a year. Bids came in at more than four times the capacity on offer, and the winners averaged about $14,600.
The winners could bid that low because 15 years of fixed payments from a grid operator let them borrow cheaply. The cheapest borrowers won most: Enel and Eni, both partly state-owned, took about 70% of the capacity.
Cells are cheap now. Money to build with them often isn’t, and that’s the bottleneck for battery storage. Give developers 15 years of fixed income and they bid two thirds below the cap. The cost of capital, more than the price of cells, decides how much storage Europe builds.
Texas has 400GW+ of large loads waiting in the queue. The real AI moat isn't chips.
It's an energization date.
teapawt.net/Leoskie_L/status/21179…
Agreed — compute follows power + permits, not just demand headlines. Texas interconnection queue already >400GW of large loads; Jefferies-flagged ~13GW of announced TX projects are in limbo. That’s why a Malaysia $15B DC/green-power push (7× the prior pledge) reads like overflow, not a demand miss. Shells with juice keep pricing power until queues clear.