North America’s data center boom is still running at full speed, but the limiting factor has shifted.
In an article published by TechFlow and attributed to Avanza Energy author Christopher Johnson, the core argument is blunt: the next phase of competition has moved away from demand and toward the physical ability to land new power supply on the grid.
The piece says data centers under construction across North America’s eight largest markets reached 7,481 MW in the first half of 2026, up 24.8% from a year earlier and above the previous record set at the end of 2024. At the same time, newly disclosed projects entering the planning pipeline have fallen for three straight quarters. Johnson argues those figures are not in conflict. They describe opposite ends of the same development pipeline, and together they show an industry rapidly consuming existing backlog while future interconnection capacity is getting squeezed.
With only about six months of prelease supply left in core U.S. hubs, the article says the hard constraint on expansion has shifted from server procurement to power availability and grid interconnection.
Construction is at a record high, while new pipeline additions are slowing
Citing CBRE, the article says data center construction across North America’s eight primary markets reached 7,481 MW in the first half of 2026, setting a new high and surpassing the previous peak recorded at the end of 2024.
Four months earlier, however, Wood Mackenzie reported a very different trend. Newly planned pipeline projects had declined for three consecutive quarters, including a 19% drop in the first quarter of 2026.
Johnson says both data sets are accurate because they measure different stages of the same pipeline. CBRE tracks projects that have already broken ground. Wood Mackenzie tracks projects newly entering the planning pipeline. If the industry is working through delayed or backlogged projects faster than it is adding new ones, a market can post record construction activity and a sharp slowdown in new project formation at the same time.
The article says that is exactly what the 2026 numbers show.
Across North America’s major markets, less than 1,500 MW of capacity remains available for prelease. At the current pace of demand, that amounts to roughly six months of supply.
CBRE Head of Data Center Research Gordon Dolven said in the firm’s first-half 2026 report: 「The next phase of growth will no longer be determined by demand, but by where new power supply can actually be delivered and interconnected.」
Two ends of the same pipeline point to the same shortage
More than 80% of all capacity now under construction has already been preleased, according to the article, up from 74.3% a year earlier. Vacancy remained at a very low 1.4%, unchanged from the end of 2025.
Atlanta overtook Northern Virginia for the first time, with nearly 2,900 MW under construction, making it the most active development market in North America. Northern Virginia still leads in total inventory and posted the highest net absorption of any market at 467.7 MW, but vacancy there has fallen to 0.2%. Every megawatt scheduled for delivery in 2026 has already been spoken for, pushing new prelease negotiations into 2027.
Pricing has moved higher as well. Asking rents rose across deployment sizes, and rents for larger tenants seeking 3 MW to 10 MW climbed 8.3%. Johnson says that is not the profile of a weak demand market. It is a market repricing around severe supply scarcity.
Wood Mackenzie’s data supports the same conclusion from another angle. Newly disclosed pipeline capacity fell to 36 GW in the first quarter of 2026, down 19% from the prior quarter and extending a slowdown that had already lasted three quarters.
But another line in the same report suggests this is not a simple cooling story. Capacity in advanced commercial negotiations, the final stage before formal agreements are signed, jumped from 37 GW to 107 GW in a single quarter. Capacity backed by signed construction contracts or power agreements rose to 195 GW, and disclosed project capital spending topped $1 trillion for the first time.
Johnson argues that a market in straightforward decline would not be maturing existing pipeline this quickly. A better reading is that older projects are moving forward while new ones are struggling to enter the queue.
A correction to earlier framing
The article also revisits two earlier claims made by Avanza Energy.
First, the firm had written in April 2026 that U.S. data center capacity under construction had posted its first year-over-year decline since 2020. Johnson now says that wording was not precise enough. CBRE’s own summary explains why: after power shortages delayed projects and briefly held back construction late last year, new starts rebounded 24.8%.
Johnson says the underlying mechanism identified earlier was still real. Power bottlenecks did delay projects from breaking ground. What changed was that previously delayed projects later moved into construction in a concentrated wave, which made the charts look like a record-setting boom.
Second, the article corrects earlier wording around PJM’s December 2025 capacity auction. The auction did clear, it says, at the $333.44 per MW-day price cap for unforced capacity, or UCAP. The real problem was that the auction did not procure enough capacity to satisfy reliability requirements, leaving a 6,623 MW shortfall.
PJM described it as the first auction in the regional transmission organization’s history, including fixed resource requirement areas, that failed to meet the minimum reliability standard. Johnson says that framing is more accurate than simply saying the market failed to clear, and more troubling.
The bottleneck is now centered on the PJM grid
The article then turns directly to PJM.
In PJM’s July 2026 capacity auction for the 2028/29 delivery year, the grid operator procured 138,318 MW of UCAP and cleared at a new price cap of $325 per MW-day. Even so, the auction still fell 6,831 MW short of the reliability requirement. PJM said: 「These auctions mark the first time in PJM’s history that the entire region failed to meet reliability standards.」
That left PJM with two consecutive reliability failures. The shortfall was 6,623 MW in December and widened to 6,831 MW seven months later, even though the second auction procured more nominal capacity under a lower price cap.
Placed side by side, the pattern is clear in the article’s telling. Both auctions produced a total procurement bill of $16.4 billion. The second auction bought more capacity, 138,318 MW versus 134,479 MW, at a lower capped price, $325 versus $333.44 per MW-day, yet the reliability gap still grew.
Johnson’s conclusion is that administrative price caps suppressed the clearing price but did nothing to solve the physical shortage of power. Supply response was limited. In the July auction, new generation and uprates cleared only 525 MW of UCAP, a small figure against a shortfall measured in the thousands of megawatts.
A 15-year backstop plan, then a delay at FERC
PJM did not wait for the next regular auction. On July 31, 2026, it filed an emergency request with the Federal Energy Regulatory Commission, or FERC, for what it called a Reliability Backstop Procurement, docket ER26-3380-000.
The proposed program would run for 15 years and would procure capacity equal to the cumulative shortfall already identified. The article quotes PJM as saying the plan was designed to address 「unprecedented large load additions, especially the rapid expansion of data centers.」 The filing set a maximum willingness to pay at a weighted average of $555 per MW-day.
Johnson says that number was not arbitrary. PJM had already run a simulation of the same auction without the price cap. In that exercise, every zone except ComEd cleared at $554.72 per MW-day, while ComEd rose to $776.69 because of transmission constraints.
PJM also cautioned that the simulation reused existing seller offers and could not predict how bidding behavior would change in a fully uncapped live auction.
The article argues that the price cap did not reduce the real financial cost of the shortfall. It merely shifted that cost out of the annual spot auction and into a 15-year special contract, with pricing that landed close to the uncapped market-clearing level.
FERC did not fully adopt PJM’s request.
On Sept. 29, FERC accepted the need for a backstop in substance but delayed the effective date by five months, to Feb. 28, 2027, subject to refund, while opening further hearings on three issues it said may be unjust and unreasonable: cost allocation, rules governing transmission owners that exit PJM, and collateral requirements imposed on load-serving entities. The article says one electric cooperative faced a collateral burden of nearly $2 billion.
PJM then withdrew the bidding window that had been scheduled to open on Sept. 30. It said it would refile a revised proposal on Oct. 29, but as of the time of writing, no new bidding schedule had been set.
Johnson writes that a mechanism meant to answer who pays for the power shortfall has itself turned into a five-month legal and administrative standoff.
The article also refers back to an Avanza Energy report from July 2026 that examined how the first shortfall turned into a $16.4 billion bill. What is happening now, Johnson says, is the next act: another large bill has arrived, repackaged as a 15-year installment plan, and even that plan is now on hold.
The project cancellation narrative is overstated
A third set of industry data, the article says, offers a more granular view of what the power shortage looks like on the ground, but that data has been simplified too aggressively in market discussion.
Currence, formerly Sightline Climate, said in a February 2026 report that among roughly 140 large projects it tracks, at least 16 GW of data center capacity had been scheduled for grid connection in 2026. Only about 5 GW of that was actually under construction. The remaining 11 GW was still in early paper-stage planning and had not broken ground.
Currence estimated that 30% to 50% of that pipeline was at risk of missing its 2026 delivery schedule. But the same report also provided a historical benchmark: in 2025, 26% of expected deliveries were delayed and another 10% were pushed back entirely.
Against that baseline, Johnson says, a 30% to 50% risk range points to a worse year, but not to a cliff-edge collapse.
SemiAnalysis, in pushing back against the widely circulated claim that half of 2026 projects had been canceled, challenged the 5 GW construction denominator itself. The firm said: 「The actual capacity under construction already far exceeds 5 GW if you count only the top two hyperscale cloud providers.」
If the denominator is understated, the share of projects deemed at risk shrinks as well. Johnson says the projects most exposed to failure are concentrated in the earliest paper-stage developments that have not yet broken ground, which lines up with the layer of the pipeline that Wood Mackenzie shows shrinking the fastest.
His conclusion is narrower than many headlines. A meaningful share of 2026 pipeline projects may miss their original schedules, but delay is not the same thing as cancellation. In his view, media coverage has often collapsed those two categories into one.
Competition is shifting from demand to deliverable supply
The article closes by putting three separate data sets together: record construction activity, a shrinking pool of newly planned projects, and two consecutive PJM auctions that failed to meet reliability standards.
Read together, Johnson says, they point to one fact. The industry is moving through existing construction backlog at a record pace, but the pool of follow-on projects is thinning. At the same time, the grid operator has twice failed to secure enough reliable supply and is trying to manage a structural deficit through long-term backstop procurement and administrative price controls.
In that framing, six months of prelease supply is not an immediate collapse. It is a countdown. The next quarter’s hard data on power supply and interconnection, not the surface appearance of construction strength, will determine where the market goes from here.
