Why Nvidia is said to be becoming a power company: The $3bn Lancium stake

$2 billion buys roughly 20% of the company upfront. A further $1 billion is contingent on Lancium hitting grid interconnection milestones. The deal values Lancium at around $10 billion.

It’s an equity stake, not project financing. Nvidia takes no construction risk and no lease obligations. It’s buying into the company that owns the land and the grid access, not the data centres sitting on top of them.

Lancium doesn’t build or run data centres itself. It develops power and land, then leases sites to operators like Crusoe, which runs the Abilene campus for Stargate. Stargate is the SoftBank, OpenAI and Oracle venture Donald Trump announced in January, pitched at up to $500 billion in AI infrastructure spend.

For the past two chip generations, Nvidia’s constraint has been fab capacity. This deal is a bet that the next constraint is watts, not wafers. Texas gives a sense of scale: ERCOT’s queue for large-load interconnection requests has swelled past 400 GW, roughly a third of all US generation capacity, and the state doesn’t have a firm figure for how much water or power its data centre pipeline will actually draw.

Governor Greg Abbott ordered a pause on new grid connections earlier this year while officials try to get a handle on it. Capacity has covered the community side of that story in Central Texas, where residents and researchers have pushed back on water use and grid strain around hyperscale sites.

Nvidia buying into Lancium turns that bottleneck into something it can price and partly control. If Abilene’s power draw slows, so does Nvidia’s own GPU deployment. A stake in the company managing the interconnection queue gives it a hand in fixing that, or at least a seat when milestones get negotiated.

Two things worth tracking from here: whether the milestone-linked second tranche becomes a template other chipmakers copy, and Lancium’s stated plan to IPO in 2027, which this cash is meant to help fund.

Data centres could use 2.7% of Texas’s water by 2030: Where they land matters