The Ledger Letter — The Chip Trade Is Priced. The Power Bill Isn’t.
Micron reports after the close. The market priced the chips. It hasn’t priced the 49-gigawatt power gap behind them.
|
The Ledger Letter
Finance Studio Advisors · Wednesday, June 24, 2026
|
|
Market Intelligence Partner
Jeff Brown picked NVIDIA before its 28,000% surge.
Now he says AI’s next trillion-dollar breakout won’t come from chips at all… but from a strange “AI Fuel” that could unlock $100 trillion in new wealth. One government-backed company sits at the center of the story. And it could see explosive growth in the coming weeks, as an urgent executive order fast-tracks demand.
Advertisement · Brownstone Research
|
The Chip Trade Is Priced. The Power Bill Isn’t.
Micron reports after the close today. The market expects 81 percent margins and $34.5 billion in revenue. The chip trade is fully priced. What is not priced is the 49-gigawatt gap between what America’s AI data centers will need by 2028 and what the grid can deliver. Morgan Stanley calls it a shortfall. The utilities call it $1.4 trillion in new capital spending. Chevron just signed a 20-year deal to power a single Microsoft data center. The bottleneck moved. It left the chip fab and landed on the power grid.
|
|
The Breakdown
Today’s disagreement: chip buyers are pricing AI demand as permanent; the power grid, the utilities, and the interconnection queue are pricing the cost of keeping it all running.
|
| 01 |
The Catalyst After the Close
Micron reports fiscal Q3 after Wednesday’s bell. Consensus expects $34.5 billion in revenue, 81 percent gross margins, and a forward guide that confirms AI memory demand is structural. The chip trade is fully priced. The question this print forces is what comes next in the supply chain.
|
|
| 02 |
The Bottleneck That Moved
US data center power consumption stands at 41 gigawatts, rivaling the combined output of every American nuclear plant. Morgan Stanley projects demand reaching 74 GW by 2028. The shortfall: 49 GW. Natural gas turbines are sold out through end of decade. Grid interconnection wait times have stretched from two years to five. The power bottleneck is no longer theoretical.
|
|
| 03 |
The Capital That Followed
US utilities have committed $1.4 trillion in new capital spending tied to AI-driven demand. Chevron signed a 20-year power deal with Microsoft for a single AI data center in West Texas. Constellation Energy is restarting Three Mile Island for Microsoft’s operations. Energy infrastructure is now attracting the capital that semiconductors absorbed two years ago.
|
|
|
|
|
By the Numbers
| US data center power (2026) | 41 GW — rivals US nuclear fleet |
| Projected demand by 2028 (MS) | 74 GW — 49 GW shortfall |
| US utility planned capex | $1.4T for AI-driven grid build |
| Grid interconnection wait | 5+ years avg — up from 2–3 |
| Utility sector value added (2024–26) | ~$500B in market cap |
Sources: Morgan Stanley Research, Deloitte, Tech Insider, CNBC, S&P Global. Figures as of June 2026.
|
|
The Full Picture
The Grid Becomes the Trade
The Chip Market Already Got Paid
Micron’s print tonight will confirm what the market already believes: AI memory demand is structural, margins are expanding, and the three suppliers who control the world’s high-bandwidth memory have pricing power for the foreseeable future. That trade is done. It has been priced into a $1.28 trillion market cap and a stock at all-time highs. The next question is not whether AI spending continues. It is where the next constraint sits.
|
|
Where the Capital Goes Next
The constraint is electricity. A single hyperscale AI training cluster draws 100 megawatts, enough to power a small city. The largest campuses under construction need 2 gigawatts. The US grid cannot deliver it on the timeline the hyperscalers need. Interconnection queues are five years deep. Gas turbines are sold out. The IEA projects natural gas and coal will meet more than 40 percent of incremental data center demand through 2030 because nothing else can be built fast enough.
That is why the capital is rotating. Chevron signed a 20-year power purchase agreement with Microsoft for a single facility in West Texas. Constellation is spending $1.6 billion to restart Three Mile Island. Tech companies have committed to financing more than 20 gigawatts of small modular reactors. In this tape, AI infrastructure is no longer a semiconductor story. It is an electricity, natural gas, nuclear, and grid-financing story. Our view: the bottleneck moved, and the capital followed it.
|
|
The Rate Hike the Grid Is Pricing
Retail electricity prices in the US have risen 42 percent since 2019. Utilities have filed $31 billion in rate increase requests. Dominion Virginia, the utility that powers the world’s largest concentration of data centers, proposed its first base-rate increase since 1992. The utility sector has added roughly $500 billion in market value over the past two years. That is not a rotation into defensives. That is the market pricing a structural demand shock in a regulated industry where the customer has no choice but to pay.
|
|
The Number That Confirms the Rotation
Worth watching: Micron’s capex guide tonight. If the company raises its spending plan above $25 billion for the fiscal year, the downstream read-through is direct. Every dollar Micron spends on new fabrication capacity is a dollar of electricity demand that the grid does not yet have. The chip market prices the revenue. The power market prices the cost. Wednesday’s print tells you how fast the second trade is arriving.
|
The chip trade made the last two years. The power trade prices the next ten. After the close tonight, the market decides whether it’s ready to look past the GPU and into the wall socket.
|
|
|
The Ledger Letter
When markets disagree, the signal is in the disagreement.
|
|
This newsletter is for informational purposes only and does not constitute investment advice. Past performance is not indicative of future results.
|
|
|
