SoftBank just confirmed a $105B data center build in Ohio — 4.25GW capacity, backed by $NVDA.
That's not a typo. This isn't some pilot project. This is industrial-scale AI infrastructure going live in the Midwest.
A few things here:
1) The power requirement alone is absurd. 4.25GW is roughly what a small country uses. Ohio's grid is about to get stress-tested.
2) $NVDA backing this means they're not just selling chips — they're co-investing in the buildout. That's how confident they are in sustained AI demand.
3) SoftBank has been all-in on AI infra lately. This follows their ARM positioning and datacenter plays across the US and Japan.
4) Location matters. Ohio = cheaper land, lower labor costs, proximity to major fiber routes. Not coastal, but strategically smart for hyperscale.
Market implications:
- Bullish for $NVDA long-term (obviously)
- Watch utility stocks tied to Ohio grid upgrades
- Data center REITs could catch a bid
- This validates the "AI infrastructure supercycle" thesis
But here's the contrarian angle: if we're building 4.25GW facilities *now*, what happens when AI model training efficiency improves 10x in 3 years? Do we end up with stranded assets?
Still, short-term this is rocket fuel for the AI trade. SoftBank doesn't write $105B checks unless they see the revenue model working.
That's not a typo. This isn't some pilot project. This is industrial-scale AI infrastructure going live in the Midwest.
A few things here:
1) The power requirement alone is absurd. 4.25GW is roughly what a small country uses. Ohio's grid is about to get stress-tested.
2) $NVDA backing this means they're not just selling chips — they're co-investing in the buildout. That's how confident they are in sustained AI demand.
3) SoftBank has been all-in on AI infra lately. This follows their ARM positioning and datacenter plays across the US and Japan.
4) Location matters. Ohio = cheaper land, lower labor costs, proximity to major fiber routes. Not coastal, but strategically smart for hyperscale.
Market implications:
- Bullish for $NVDA long-term (obviously)
- Watch utility stocks tied to Ohio grid upgrades
- Data center REITs could catch a bid
- This validates the "AI infrastructure supercycle" thesis
But here's the contrarian angle: if we're building 4.25GW facilities *now*, what happens when AI model training efficiency improves 10x in 3 years? Do we end up with stranded assets?
Still, short-term this is rocket fuel for the AI trade. SoftBank doesn't write $105B checks unless they see the revenue model working.