AI Infrastructure Is Being Repriced in Megawatts, Not Acres
DN compares Paradex, OKX, Kraken, Hyperliquid, Aster, GRVT, Bybit, Coinbase and Binance for AI trading agents across credentials, APIs, state certainty, emergency controls and autonomous execution.
The Megawatt Basis: Why Grid Access Could Become More Valuable Than the Data Center
AI infrastructure is still discussed in terms of chips, buildings and capital. But the asset deciding when billions of dollars can begin earning revenue may be much simpler: a documented megawatt that can actually be delivered on time.
The Signal
- AI data centers are turning reliable, near-term electricity access into a distinct economic asset. The scarcity is increasingly not land, capital or even chips. It is deliverable power on a credible date.
- Powered land in primary US data-center markets averaged roughly $584,000 per MW in 2026 year-to-date, about 51% higher than a year earlier, according to Cushman & Wakefield.
- The same research estimates modern greenfield data-center development at about $17.6 million per MW. That means the observable powered-land premium can look small relative to total project cost while controlling when the rest of the capital becomes productive.
- PG&E reported more than 12.7 GW of data-center pipeline demand in June 2026, but only 490 MW had reached an executed Interconnection Construction Agreement and 140 MW was under construction.
- FERC has ordered all six federally regulated regional grid operators to justify or reform how large loads such as data centers connect to the transmission system. Regulators now explicitly describe speed-to-power as an economic and national-security issue.
- DN calls the difference between an ordinary site's real-estate value and the economic value of documented, near-term capacity the Megawatt Basis.
- The proprietary tool below calculates not only a powered-land premium, but also Time-to-Power Option Value, Energization Alpha, Stranded Shell Exposure and Hardware-Cycle Erosion.
Every artificial-intelligence model has a compute problem.
Every compute problem eventually has a power problem.
And every power problem has a deceptively simple question:
When can the electricity actually arrive?
That question is beginning to reprice the data-center market.
The industry was once analyzed like specialized real estate.
Land mattered.
Location mattered.
Fibre mattered.
Buildings mattered.
Tenant quality mattered.
All still do.
But AI changes the hierarchy.
A perfectly designed data center without power is not a productive computing asset.
It is a very expensive building waiting for permission to become one.
The Market Is Starting to Price Megawatts Instead of Acres
The shift is becoming visible in property markets.
Cushman & Wakefield's 2026 Data Center Development Cost Guide says power is now the gating factor for development.
Powered land in primary US data-center markets has averaged roughly $584,000 per MW in 2026 year-to-date.
That is approximately 51% above the previous year and 35% above the five-year average reported by the firm.
The change is important because land is no longer being valued only in acres.
It is increasingly being valued in:
megawatts.
A 100-acre site with no credible power date may be less valuable to an AI developer than a much smaller parcel with signed access to several hundred megawatts.
Real estate is becoming energy infrastructure.
The key AI infrastructure real-estate metric is shifting from price per acre toward value per verified deliverable megawatt. Location still matters, but power certainty increasingly determines whether the location can be monetized at all.
The Visible Power Premium Is Only the Beginning
There is an important trap in the numbers.
Cushman & Wakefield estimates modern greenfield data-center development at roughly $17.6 million per MW.
Powered land at $584,000 per MW is only a small fraction of that headline development cost.
It can therefore appear that the land premium is economically minor.
That interpretation misses the option value.
The power connection determines when the other $17 million or more per MW can start generating revenue.
Consider an illustrative 100 MW development.
At $17.6 million per MW, total development cost would be roughly $1.76 billion.
Suppose Site A can energize two years earlier than Site B.
Even before considering revenue, an 8% annual carrying cost on $1.76 billion equates to roughly $282 million across two years.
The precise economics will differ dramatically by structure and when capital is actually deployed.
But the principle is clear.
A few tens of millions of dollars paid to secure earlier power can potentially influence hundreds of millions of dollars of financing carry and foregone operating economics.
Time-to-Power Is an Embedded Financial Option
This is why DN believes power availability should not be treated only as a utility constraint.
It is an option.
A site that can energize earlier gives the developer the right to begin converting:
- land into revenue,
- buildings into revenue,
- GPUs into revenue,
- customer contracts into revenue,
- and debt into productive leverage.
The value of that option rises when:
- AI demand is strong,
- hardware is expensive,
- financing costs are high,
- grid queues are long,
- and compute generations are turning over quickly.
DN calls this:
Time-to-Power Option Value.
A secured megawatt is not merely energy capacity. It is an option on earlier compute revenue. The longer the alternative energization delay, the more valuable the option becomes.
The AI Depreciation Trap Makes Delay More Expensive
This connects directly to the previous DN research on the AI Depreciation Trap.
Modern AI hardware can have a materially shorter economic cycle than the buildings housing it.
Suppose a developer orders hardware based on an expected four-year economic refresh cycle.
If grid delays push energization back by 24 months, half of that assumed cycle has passed before the site reaches production.
The hardware may still be physically new.
The competitive benchmark has moved.
New accelerator generations may offer:
- higher performance per watt,
- greater memory bandwidth,
- better interconnect,
- lower inference costs,
- and improved rack density.
The grid delay can therefore create a technology problem before the data center even opens.
DN calls this:
Hardware-Cycle Erosion.
The Building Can Become the Stranded Asset
Conventional thinking treats the building as the durable asset and the technology as the depreciating asset.
That remains broadly correct.
But construction sequencing creates another danger.
Capital can be committed to:
- land,
- site preparation,
- buildings,
- cooling,
- electrical equipment,
- networking,
- and hardware
before full power certainty exists.
If energization is delayed or materially resized, a portion of that capital can sit idle.
The result is what DN calls:
Stranded Shell Exposure.
It is the amount of pre-energization capital whose economic productivity depends on power arriving as expected.
PG&E Shows Why Pipeline Megawatts Are Not Real Megawatts
One of the most revealing pieces of 2026 utility data comes from PG&E.
The company reported a data-center pipeline of approximately 12.71 GW in June.
But those megawatts sat at radically different stages of development.
Roughly:
- 8.2 GW remained in application and preliminary engineering,
- 3.88 GW had reached final engineering,
- 490 MW had an executed Interconnection Construction Agreement,
- and only 140 MW was in construction.
That is a huge difference between:
requested power
and
contractually advanced power.
The market often discusses data-center pipelines as if all megawatts are equal.
They are not.
AI infrastructure analysis needs to separate claimed megawatts from verified megawatts. Pipeline capacity should be discounted according to the documentary and construction stage supporting it.
The DN Verified Megawatt Ladder
DN proposes a simple evidence hierarchy for power claims.
The Difference Could Matter to Valuation
Imagine two AI infrastructure companies each announce a 2 GW development pipeline.
Company A has:
- site control,
- executed power agreements,
- equipment procurement underway,
- and a defined energization schedule.
Company B has:
- land options,
- utility applications,
- preliminary discussions,
- and no construction agreement.
Both can report 2 GW.
Economically they do not own the same thing.
Pipeline megawatts therefore need a confidence haircut in the same way analysts discount unproven reserves, speculative property developments or early-stage drug pipelines.
The Grid Queue Is Becoming an Economic Asset
The value of an advanced interconnection position comes partly from how difficult the alternative has become.
Berkeley Lab's 2026 generation-interconnection dataset shows that power projects reaching commercial operation in 2025 spent more than five years at the median between entering the interconnection process and operating.
That dataset covers generation rather than large-load interconnection, so it should not be treated as a direct measure of data-center waiting times.
But it illustrates the larger constraint.
Electricity demand can be proposed faster than the generation and transmission infrastructure required to serve it can be built.
The physical grid is becoming the pacing mechanism for digital growth.
FERC Is Rewriting the Rules Because the Old Process Is Too Slow
The regulatory response is now explicit.
In June 2026, the Federal Energy Regulatory Commission ordered all six regional grid operators under its jurisdiction to justify or reform rules governing the connection of data centers and other large loads.
FERC describes these new loads as unusually:
- large,
- concentrated,
- fast-growing,
- and capable of rapidly changing electricity consumption.
For policy purposes, the current proceeding generally treats loads above 20 MW as large loads.
That is tiny compared with the scale of proposed AI campuses.
Some announced projects are measured in hundreds of megawatts or multiple gigawatts.
The interconnection process is therefore moving from a back-office infrastructure question into an AI competitiveness question.
Speed-to-Power Has Become Policy
FERC's language is revealing.
The regulator now explicitly talks about:
speed-to-power.
That means time-to-power has graduated from a developer inconvenience into a variable being considered at the highest level of US energy regulation.
The Commission is examining:
- faster study processes,
- large-load cost recovery,
- flexible transmission arrangements,
- co-location,
- load curtailment,
- electrically proximate generation,
- and bring-your-own-new-generation structures.
Those are not merely grid reforms.
They are attempts to manufacture faster megawatts.
As grid access becomes scarce, regulation itself becomes part of asset value. The valuable AI site is increasingly not just land + power, but land + power + permission + time certainty.
The Real Asset Is Power Plus Permissions
A power plant can exist without being able to serve a particular data center immediately.
Transmission may be constrained.
The substation may need upgrading.
Transformers may not be available.
The load study may be incomplete.
Permits may remain unresolved.
Cost allocation may be disputed.
This is why headline national generation capacity tells investors less than they might assume.
The useful resource is not electricity in the abstract.
It is electricity:
at the correct location, at the required scale, with sufficient reliability, through infrastructure that exists, under legal arrangements that permit delivery, on a commercially useful date.
The Transformer Is Becoming Part of the AI Supply Chain
The grid constraint extends deep into manufacturing.
Reuters reported in July that lead times for some high-voltage transformers had reached approximately 160 weeks, compared with about 143 weeks in 2024.
Circuit breakers and switchgear are also under pressure.
This matters because AI infrastructure discussions often focus on semiconductor supply chains.
Yet a shortage of a transformer can strand GPUs just as effectively as a shortage of GPUs can strand a data-center rack.
The AI supply chain therefore extends from:
silicon
to
substations.
AI Infrastructure Is Migrating Toward Electricity
This is already changing geography.
Reuters, citing JLL data, reported that the average European data-center site planned for 2026 to 2028 is about 175 kilometres from a major city.
Projects built from 2022 to 2025 averaged around 46 kilometres.
The reason is straightforward.
AI workloads can often tolerate greater distance from dense urban centers than latency-sensitive traditional applications.
Developers can therefore move toward:
- cheaper land,
- better power availability,
- generation resources,
- and faster interconnection.
That changes the location premium.
AI is changing data-center geography from compute near users toward compute near energy. Where latency requirements permit, electrons can become more important than postcode.
The Price Dispersion Is Extraordinary
The same European research illustrates how dramatically power scarcity can alter site economics.
Powered land has been quoted around €2.7 million per MW in constrained Amsterdam compared with roughly €200,000 per MW in tertiary markets such as Bordeaux.
That is not simply a real-estate spread.
It reflects:
- grid scarcity,
- demand concentration,
- permitting,
- network capacity,
- and time-to-power.
Two megawatts of electricity can deliver the same unit of physical power while having radically different economic prices.
This is the essence of the Megawatt Basis.
What Is the Megawatt Basis?
In commodity markets, a basis represents the difference between related prices across locations, qualities or delivery structures.
DN applies similar thinking to electricity access for AI infrastructure.
The Megawatt Basis is the economic premium attached to a verified, deliverable megawatt relative to the underlying unpowered site.
But a proper Megawatt Basis should capture more than property price.
It includes:
- the explicit powered-land premium,
- time-to-power advantage,
- financing carry avoided,
- revenue brought forward,
- grid-upgrade cost,
- power-price differences,
- delivery certainty,
- and the effect of delay on the technology cycle.
The Cheapest Power May Not Produce the Highest Return
Investors naturally focus on electricity price.
That is important.
A hyperscale facility consumes enormous amounts of energy.
But a site offering electricity $10 per MWh cheaper five years from now may be economically inferior to a more expensive site that can operate next year.
The correct calculation compares:
power price
against
time-to-power.
A developer should rationally pay more for electricity if the extra cost is smaller than the value of earlier revenue and avoided financing carry.
This creates:
Energization Alpha.
The economically cheapest megawatt is not necessarily the megawatt with the lowest electricity price. It is the megawatt with the lowest total cost of waiting, connecting, financing and operating.
The AI Race Could Become an Energy-Arbitrage Race
Companies can respond to grid scarcity in several ways.
They can move.
They can wait.
They can co-locate with generation.
They can build generation.
They can accept curtailment.
They can sign long-term power contracts.
They can use batteries.
They can develop natural-gas generation.
They can contract nuclear output.
They can invest in transmission.
They can build in a different country.
This is why AI infrastructure increasingly looks less like software and more like industrial strategy.
Google's Finland Deal Shows the Scale of the Shift
In September, Google announced a major Finnish AI infrastructure investment connected to a long-term power arrangement with the Loviisa nuclear plant.
The agreement covers up to half of the plant's output for 22 years.
Whatever view one takes of the local political debate around electricity availability, the structure is revealing.
A technology company is securing decades of power supply to support the AI stack.
The compute business is reaching upstream into energy.
South Korea Shows the National Scale
South Korea offers another example.
Government estimates suggest AI data centers and semiconductor expansion could add roughly 25 to 30 GW to national electricity demand.
That is power-system-scale growth.
The country is therefore debating not only AI policy but also nuclear capacity and long-term electricity planning.
AI strategy becomes energy strategy.
The Grid Is Becoming Part of AI Sovereignty
A country can possess:
- excellent AI researchers,
- capital,
- cloud companies,
- semiconductor access,
- and supportive regulation
and still fail to build enough compute if electricity cannot be delivered.
That means compute sovereignty increasingly depends on:
- generation,
- transmission,
- transformers,
- permitting,
- grid flexibility,
- and interconnection governance.
AI sovereignty is therefore partly grid sovereignty.
The New Collateral Hierarchy
The previous DN AI Depreciation Trap research argued that power access could prove more durable collateral than the accelerator.
The Megawatt Basis extends that thesis.
Consider the hierarchy:
| Asset | Economic life | Main risk | Potential durability |
|---|---|---|---|
| AI accelerator | Short | Technological obsolescence | Low to moderate |
| Server / rack infrastructure | Short to medium | Hardware refresh | Moderate |
| Data-center shell | Long | Power or tenant mismatch | High if reusable |
| Cooling / electrical infrastructure | Medium to long | Density changes | Potentially high |
| Fibre connectivity | Long | Route / capacity economics | High |
| Documented grid access | Potentially long | Regulatory, delivery and tariff risk | Very high where capacity is scarce |
| Low-cost dispatchable power | Long | Commodity, policy and generation risk | Potentially very high |
The Megawatt May Become the Real Estate
This produces a deeper conclusion.
In traditional real estate, the building sits on the land.
In AI infrastructure, economic value increasingly sits on the power allocation.
The land is where the right is exercised.
The building converts it into compute.
The GPU converts it into intelligence.
If power access becomes the hardest component to reproduce, the megawatt begins to behave like the scarce underlying asset.
AI infrastructure may invert traditional real-estate economics. The land does not merely host the valuable asset. Increasingly, the site is valuable because it carries a credible claim on scarce electricity capacity.
The Speculative Megawatt Problem
Scarcity creates incentives to claim capacity early.
Developers can seek grid studies for projects that may never be built.
FERC has explicitly identified speculative and duplicative large-load requests as a problem.
A developer can effectively shop multiple locations seeking the fastest and cheapest connection.
The same underlying demand can therefore appear in several planning pipelines.
This can:
- inflate load forecasts,
- consume utility engineering resources,
- distort investment signals,
- encourage unnecessary grid upgrades,
- and confuse investors assessing pipeline quality.
This is why site control, deposits, engineering milestones and contractual commitments matter.
The Megawatt Basis should rise with evidence quality, not marketing volume.
The Power Certainty Discount
DN proposes applying a Power Certainty Discount to announced capacity.
A 500 MW campus in construction is not economically equivalent to a 500 MW concept with a utility application.
Analysts should discount capacity according to factors such as:
- physical site control,
- utility study stage,
- executed agreements,
- upgrade funding,
- equipment procurement,
- generation support,
- construction status,
- and expected energization date.
This is particularly important when comparing AI infrastructure companies whose valuations depend heavily on future MW pipelines.
Who Could Capture the Megawatt Basis?
The obvious beneficiaries are not limited to data-center landlords.
Utilities
Large new loads can expand rate bases and spread fixed system costs if tariffs allocate expenses appropriately.
Transmission developers
Grid scarcity increases the value of moving electricity from generation-rich regions to compute demand.
Transformer and switchgear manufacturers
Long equipment lead times move electrical hardware into the AI bottleneck stack.
Power developers
Natural gas, nuclear, geothermal, renewables and storage can all gain from hyperscaler demand depending on location and reliability requirements.
Powered-land developers
Developers capable of advancing land through grid, planning and permitting milestones can potentially manufacture higher-value real-estate inventory.
Flexible-load software
Data centers able to modulate demand may gain access to faster or cheaper connection structures.
Energy-rich countries and regions
AI workloads that tolerate geographic distance can migrate toward abundant electricity.
The Biggest Risk: Paying for Power That Never Arrives
Every scarcity premium can become a bubble.
Powered land is no exception.
A site marketed as having access to power may ultimately face:
- network upgrades,
- new generation requirements,
- permitting delays,
- transformer shortages,
- tariff changes,
- community opposition,
- curtailment conditions,
- or a later energization date.
The premium should therefore attach to documentation and delivery probability.
Not the word:
"powered."
The next data-center mispricing may occur between marketing megawatts and contractual megawatts. Investors should diligence the paperwork behind every claimed MW with the same skepticism applied to reserves, customer contracts or backlog.
DN Megawatt Basis Monitor
Signals Worth Tracking
Track the infrastructure behind the AI trade
AI infrastructure increasingly connects semiconductors, utilities, nuclear, natural gas, grid equipment, data centers and credit markets. TradingView can be used to monitor cross-market price trends and company performance, while ASCN provides AI-assisted digital-asset and technology research workflows. These are affiliate links.
DN Megawatt Basis Engine
Traditional real-estate valuation struggles with powered land because the most important variable may not be what sits on the parcel today.
It is how much sooner that parcel can begin producing digital output.
The engine below estimates the economic difference between an earlier powered site and a slower alternative.
Megawatt Basis & Time-to-Power Engine
Estimate powered-land premium, financing carry avoided, revenue brought forward, energy-cost trade-offs, Stranded Shell Exposure and Hardware-Cycle Erosion. Defaults are illustrative and should be replaced with project-specific assumptions.
Calculating...
How the Megawatt Basis Engine Works
The tool separates four kinds of value that are often mixed together.
1. Explicit powered-land premium
What does the site cost because it carries documented power access?
2. Time-to-Power Option Value
How much operating value and financing carry may be preserved by reaching energization earlier?
3. Operating penalty
Does the faster site require a higher electricity price or expensive network upgrades?
4. Certainty
How much confidence should be attached to the claimed energization path?
These variables create an effective Megawatt Basis.
The purpose is not to declare one site cheaper.
It is to reveal the economic value of time hidden behind the utility connection.
A Counterintuitive Example
Suppose a slower site offers electricity $10 per MWh cheaper.
That sounds attractive.
But suppose the cheaper site takes four additional years to energize.
During those four years:
- capital carries interest,
- revenue is delayed,
- customer contracts may move elsewhere,
- construction inflation continues,
- and the compute architecture may change.
The electricity saving can be real and still be economically dominated by the waiting cost.
This is why the correct AI infrastructure question is not:
Where is power cheapest?
It is:
Where is verified power cheapest after including time?
What Would Prove the Thesis Wrong?
The Megawatt Basis thesis weakens materially if:
- grid interconnection times fall sharply across major data-center markets,
- transformer and switchgear lead times normalize,
- large amounts of dispatchable generation come online faster than AI demand,
- data-center power demand forecasts are materially overstated,
- AI efficiency improvements reduce electricity demand faster than workload growth increases it,
- powered-land premiums fall while delivery certainty improves,
- large-load reforms make power access standardized and rapidly available,
- and portable or modular generation makes utility interconnection much less important.
Any combination of these could reduce the scarcity premium.
There is also a more fundamental challenge.
If the AI investment cycle slows sharply, planned power demand could be cancelled faster than grid infrastructure can be repurposed.
That would reverse the scarcity trade.
Some highly priced powered land could then prove overvalued.
The Bull Case Has Its Own Constraint
The stronger the AI demand thesis becomes, the more valuable power access becomes.
But that eventually creates a circular problem.
Rising power scarcity increases:
- development cost,
- utility investment,
- community opposition,
- energy prices,
- and political scrutiny.
The infrastructure needed to satisfy AI demand can itself raise the cost of AI.
That means energy is not merely an external input.
It can become the marginal governor on the compute supercycle.
The Real AI Bottleneck May Keep Moving Down the Stack
First the market worried about GPUs.
Then advanced packaging.
Then high-bandwidth memory.
Then data-center capacity.
Now the constraint is increasingly:
power,
transmission,
transformers,
permitting,
and physical construction.
The AI boom keeps moving the bottleneck toward slower industries.
That is important because software scales quickly.
Infrastructure does not.
AI's largest economic bottleneck may ultimately be created by the mismatch between the exponential scaling culture of software and the linear construction timelines of energy infrastructure.
The Bigger Conclusion
A data center can be copied.
A server can be purchased.
A GPU can be replaced.
Capital can be raised.
Land can often be found.
But a large block of reliable electricity at the correct location, delivered through the correct infrastructure on a useful date, cannot always be manufactured quickly.
That changes what investors should value.
In the next phase of the AI buildout, the most important question may not be:
How many data centers does this company control?
It may be:
How many verified megawatts can it actually energize, and when?
That is a more difficult number to market.
It is also a much more useful number to own.
AI has made intelligence increasingly dependent on industrial infrastructure.
The data center is where compute lives.
But the grid connection is what allows it to become valuable.
In a world racing for artificial intelligence, the scarce asset may therefore be surprisingly primitive.
A megawatt.
At the right place.
At the right time.
Primary Sources & Evidence
- International Energy Agency, Energy and AI, electricity-demand and data-center projections.
- US Department of Energy and Lawrence Berkeley National Laboratory, 2024 United States Data Center Energy Usage Report.
- US Department of Energy, National Transmission Needs Study, Draft 2026.
- Federal Energy Regulatory Commission, Large Load Interconnection Proceedings and June 2026 Show Cause Orders.
- Federal Energy Regulatory Commission, Commissioner David Rosner's June 2026 remarks on large-load interconnection.
- Lawrence Berkeley National Laboratory, Queued Up: 2026 Edition.
- PG&E Corporation, Second Quarter 2026 Earnings Presentation.
- Cushman & Wakefield, 2026 Data Center Development Cost Guide.
- Reuters, Europe AI data centres seek cheaper, quicker energy and land, August 2026.
- Reuters, US power companies scramble to secure equipment as surging data center demand strains supplies, July 2026.
- Reuters, US power use to beat record highs in 2026 and 2027 as AI use surges, September 2026.
- Reuters, Finland risks strained power supply after Google AI deal, September 2026.
- Reuters, South Korea's power demand set to soar on AI boom, September 2026.
Frequently Asked Questions
What is the Megawatt Basis?
The Megawatt Basis is a Decentralised News framework for estimating the economic premium attached to documented, deliverable electricity capacity at a data-center site relative to an unpowered or slower-to-energize alternative. It incorporates land premium, time-to-power, financing carry, operating value, grid costs and electricity-price differences.
What is powered land?
Powered land generally refers to land with credible access to sufficient electricity capacity for a planned development. The strength of that claim can vary dramatically depending on utility studies, executed agreements, upgrade requirements, construction progress and actual energization.
Why is power so important to AI data centers?
AI accelerators consume significant electricity at high densities. A facility cannot monetize installed compute without sufficient reliable power, making utility capacity, substations, transformers and generation increasingly important parts of the AI infrastructure stack.
What is Time-to-Power Option Value?
Time-to-Power Option Value is DN terminology for the economic value created by energizing a project earlier, including operating contribution brought forward and financing carry potentially avoided during the period saved.
What is Hardware-Cycle Erosion?
Hardware-Cycle Erosion measures how much of an assumed compute refresh cycle passes while a project waits for power. Long energization delays can cause a meaningful portion of a GPU generation's competitive economic life to pass before the facility begins operating.
Are all announced data-center megawatts equal?
No. Announced capacity can range from an early development concept to contracted capacity already under construction. DN's Verified Megawatt Ladder separates marketing, applied, engineered, contracted, construction and energized capacity.
Why could a more expensive electricity market still be attractive?
A site with higher electricity costs may still have better economics if it can energize materially sooner. Earlier revenue and lower financing carry can outweigh part of the long-term energy-price disadvantage.
Could powered land become overvalued?
Yes. Power claims can fail, projects can be delayed, AI demand can weaken and electricity-market conditions can change. Investors should verify utility documentation, cost responsibility, delivery dates and construction milestones rather than relying solely on the label "powered land."
What could reduce the value of grid access?
Faster interconnection, abundant new generation, improved grid infrastructure, shorter equipment lead times, more flexible loads or weaker AI electricity demand could all reduce the scarcity premium attached to powered sites.
