Chapter 8: Legacy Supply Chain Constraints
- Sean M. Walsh
- Dec 11, 2025
- 12 min read
Updated: Jan 23
The legacy system supply chain is wildly insufficient to meet the exploding electricity demand by American computing.
For want of a nail, the shoe was lost.
For want of a shoe, the horse was lost.
For want of a horse, the rider was lost.
For want of a rider, the battle was lost.
For want of a battle, the kingdom was lost.
All for the want of a horseshoe nail.
~ 17th Century English Proverb, popularized by Benjamin Franklin, Poor Richard's Almanac (1758)
Key Points:
1. Supply chains for legacy datacenter electricity supply systems cannot scale fast enough to meet demand by datacenters over the next 5-10 years, and probably longer.
2. This underscores the opportunity... a datacenter electricity system that avoids these bottlenecks can disrupt the industry and capture enormous profits.
3. Off-grid Solar Computing Clusters are just such a system, as they are comprised of much smaller, easier-to-manufacture components from alternate supply chains that are far larger and more diversified.
Key Stats:
5-7 years - current manufacturing lead times for high-voltage electric transformers (required for legacy datacenter projects).
2x - approximate increase in high-voltage electric transformer prices from 2021-2024.
80% - percentage of high-voltage electric transformers now imported into the US, mostly from China.
200 - number of high-voltage electric transformers, imported from China, verified to contain back-door remote kill switches that were secretly added by Chinese manufacturers.
80,000 types - number of different types of electric transformers currently in use across the US electric grid.
75,000 - number of substations comprising the US electric grid, with an average transformer age of 40 years, which is approximately 10 years beyond the average design life.
6-7 years - average natural gas turbine lead times, despite the highest US order volume in over 20 years.
500,000 electricians - estimated number of new electricians for the US electric grid that must be trained and hired by the mid-2030s.
2035 - estimated year by which nuclear power plants, including SMRs, will make a meaningful impact on total US electricity production.
For Want of a Nail
On August 22, 1485, King Richard III of England charged headlong into the Battle of Bosworth Field. His horse stumbled, threw him, and he was killed — ending the Plantagenet dynasty and ushering in the Tudors. Legend holds that Richard's horse had lost a shoe because the blacksmith lacked a single nail.
Five centuries later, the proverb remains instructive: small constraints cascade into systemic failures.
In September 2020, Ford Motor Company began parking thousands of brand-new F-150 trucks at Kentucky Speedway's parking lot. This was not because the trucks were defective — they were nearly complete. They were missing a single component: a semiconductor chip worth roughly $1. Without that chip, a $55,000+ truck couldn't move. By 2021, Toyota had cut worldwide production by 40%. General Motors halted production at almost all North American plants. Across the industry, the global chip shortage caused the loss of more than 35 million vehicles and over $200 billion in revenue.
In March 2021, the container ship Ever Given lodged itself sideways in the Suez Canal — a 400-meter vessel wedged into a 265-meter channel. For six days, 432 vessels waited, carrying $92.7 billion in cargo. The blockage disrupted $9.6 billion in global trade per day. The ripple effects lasted months.
These dramatic disruptions pale compared to the structural supply chain constraints now strangling the data center industry. Every key component required to power AI infrastructure — transformers, turbines, nuclear reactors, and the electricians to install them — faces bottlenecks measured in years, not weeks.
The Transformer Crisis - An Industry on the Brink
The headline from Power Magazine captured it perfectly: "The Transformer Crisis: An Industry on the Brink."
Transformers are the backbone of the electric grid — the devices that step voltage up for long-distance transmission and back down for distribution to homes and businesses. Without them, electricity cannot flow. And the United States is running dangerously short.
According to Wood Mackenzie's August 2025 analysis, power transformers will face a 30% supply deficit in 2025, with distribution transformers trailing at a 10% shortfall. Since 2019, demand for power transformers has surged 116%, while distribution transformer demand has risen 41%. Lead times for large power transformers now stretch from 120 to 210 weeks — up to four years. Some facilities report five-year waits for new orders.
Richard Voorberg, President of Siemens Energy North America, offered a sobering assessment: "I've been in this industry for 32 years, and I've seen booms and busts. I've never seen something like this that is going up at such a rate..."
The United States has more than 80,000 different transformer types in service across its grid. Pre-pandemic, lead times ran 30-60 weeks. Now manufacturers are booking orders into 2029 and beyond. And prices have exploded: generation step-up transformers are up 45% since 2020, power transformers up 77%, and distribution transformers up 78-95% depending on specification.
The root causes are multiple and reinforcing. Electrical steel (GOES) prices have nearly doubled. Copper has increased over 40%. But the constraint isn't just materials — it's manufacturing capacity. As Voorberg explained: "It's skilled labor — there's a lot of manual labor in the windings of a transformer."
OEMs have announced $1.8 billion in North American capacity expansions since 2023. But Wood Mackenzie is clear: it won't be enough. "The convergence of accelerating electricity demand, aging infrastructure, and supply chain vulnerabilities has created constraints that will persist well into the 2030s."
As Ben Boucher, Wood Mackenzie's Senior Analyst, summarized: "The U.S. transformer market stands at a critical juncture, with supply constraints threatening to undermine the nation's energy transition and grid reliability goals."
The Trojan Horse - Chinese Backdoors Into The US Grid
The supply crisis carries an even darker dimension: national security.
In the summer of 2019, U.S. federal authorities seized a 500,000-pound Chinese transformer at the Port of Houston. Originally destined for a Colorado substation operated by the Western Area Power Administration, the massive device was quietly diverted under national security escort to Sandia National Laboratories in New Mexico.
What they found was stunning.
According to Latham Saddler, former Director of Intelligence Programs at the National Security Council: "They found hardware that was put into [the transformer] that had the ability for somebody in China to switch it off."
This wasn't speculation or theoretical vulnerability. It was confirmed hardware — a backdoor deliberately installed that would allow remote access from Beijing. The manufacturer was JiangSu HuaPeng Transformer Co., Ltd. (JSHP), a Chinese company whose website boasts that its transformers handle 20% of the electrical load for Las Vegas and 10% of the load for New York City.
In May 2020, President Trump signed Executive Order 13920, declaring a grid security emergency and banning acquisition of bulk-power equipment from foreign adversaries. In January 2021, the Biden administration suspended that order on day one. From 2020 to 2023, imports of extra-large Chinese transformers continued.
The Director of National Intelligence's 2021 report stated plainly: "China is the world's leading supplier of advanced grid components for ultra-high-voltage systems, such as transformers, circuit breakers, and inverters, which we assess creates cyber vulnerability risks."
At least 200 high-voltage Chinese transformers are now embedded throughout the U.S. grid. Some supply power to critical military installations. Some route electricity to major American cities. All represent potential attack vectors in any future conflict.
As Leon Panetta warned in 2012: "The collective result of these kinds of attacks could be a cyber Pearl Harbor. An attack that would cause physical destruction and the loss of life."
The transformer shortage has created a disturbing choice: wait years for domestic supply, or import from an adversary who has demonstrated both capability and intent to compromise our infrastructure.
The Turbine Bottleneck
Natural gas power plants are supposed to be the reliable backup — firm, dispatchable power available when wind doesn't blow and sun doesn't shine. But gas turbines face their own supply chain crisis.
U.S. gas turbine orders hit 14 GW in 2024 — the highest level since 2001. Lead times have stretched from 3-4 years to 6-7 years. In January 2025, Engie withdrew from Texas Energy Fund projects, citing equipment procurement constraints.
Only three OEMs dominate the global market: GE Vernova, Mitsubishi Power, and Siemens Energy. All are operating at full capacity. Siemens recently reported a record $158 billion backlog for natural gas turbines, with some turbine frames sold out for seven years.
The manufacturers are reluctant to expand. They've been burned before by boom-bust cycles in power generation. They face the same skilled labor shortages affecting transformer production. And they know that utility planning horizons are notoriously fickle — the plant ordered today might be canceled tomorrow if policy shifts.
According to Deloitte's 2025 analysis: "Gas-based generation projects that haven't already contracted equipment aren't expected to become available until the 2030s."
Project costs have escalated to approximately $2,000 per kilowatt — and that's before accounting for tariffs on imported components. For data centers needing power now, the turbine bottleneck represents another multi-year delay with no workaround.
The Nuclear Mirage
Small Modular Reactors (SMRs) are the great hope of the nuclear renaissance. Compact, factory-built, supposedly cheaper than traditional nuclear plants — they've attracted billions in investment and captured the imagination of Big Tech. Google, Microsoft, Amazon, and Meta have all announced nuclear exploration initiatives.
There's just one problem: SMRs don't exist at commercial scale.
As of 2025, only two commercial SMRs operate anywhere in the world — one in China, one in Russia. Only three SMR designs have been approved by the U.S. Nuclear Regulatory Commission. And the flagship American SMR project — NuScale's Carbon Free Power Project in Idaho — collapsed spectacularly in November 2023.
The NuScale story is instructive. Originally scheduled to deliver 720 MW by 2023 for $3.6 billion, the project was repeatedly scaled back and delayed. By 2023, it had shrunk to 462 MW with an operational date of 2029 — and costs had ballooned to $9.3 billion, a 158% increase. The projected electricity price jumped from $55 per megawatt-hour to $89 — roughly three times the cost of utility-scale solar with storage. Utah municipalities backed out, and the project was terminated.
The Information Technology and Innovation Foundation's April 2025 assessment was blunt: "Investments and hype are fueled by the promise of SMRs, not the current reality. Unlike large reactors, initially high SMR costs may fall because they are designed to be built — partly or completely — in a factory. But we don't yet know whether SMRs will crack the scale-up problem; that question cannot be answered for at least a decade."
Even optimistic timelines place the first wave of commercial SMR deployments well into the 2030s, with advanced Generation IV designs not arriving until 2040 or later. For data centers that need power now — not in a decade — nuclear is a dangerous mirage on the horizon.
The Missing Half-Million
Every primary electrical component discussed so far — transformers, turbines, reactors — requires one common resource to deploy: specially trained electricians.
When Microsoft executives met with White House officials in late 2024 to discuss AI infrastructure challenges, they were asked to identify the single biggest obstacle to data center expansion. The expected answers were permitting delays, grid interconnection queues, or transformer shortages.
Instead, they highlighted people.
"One of the most striking aspects of a data center construction site is the large number of electricians at work," Microsoft noted in their analysis. "Due to local labor shortages, we are employing some electricians who are commuting as far as 75 miles away or even temporarily relocating from across the country to fill these critical roles."
The numbers are stark. The United States currently has approximately 780,000 active electricians, with 80,000 job openings. Nearly 30% of union electricians are between the ages of 50 and 70, according to the National Electrical Contractors Association. It's reasonable to project 20,000 electricians retiring each year — 200,000 over the next decade.
Meanwhile, demand is surging. Microsoft estimates that meeting growing electricity needs will require adding 30,000 electricians annually — 300,000 over the decade, on top of the 200,000 retirees who must be replaced.
Total: 500,000 new electricians needed by the mid-2030s.
We don't have a national strategy to recruit and train them. The electrical workforce is projected to shrink 14% by 2030, while demand could increase 25% over the same period. In California, there's one certified electrician for every 478 households. You can't learn electrical work remotely through online courses — it requires hands-on training that the pandemic disrupted for years.
Every grid project is competing for the same finite pool of specialty workers. The electrician shortage is a crucial bottleneck: even if transformers and turbines magically appeared tomorrow, there wouldn't be enough hands to install them.
The Seven Gaps To Powering AI
In June 2025, Deloitte released its AI Infrastructure Survey, polling 120 executives from U.S.-based power and data center companies. Their findings paint a comprehensive picture of an industry hitting walls in every direction.
AI data center power demand will grow from 4 GW in 2024 to 123 GW by 2035 — a 30-fold increase. While AI accounted for only 12% of data center demand in 2024, it could represent 70% of the projected 176 GW total by 2035.
Deloitte identified seven gaps throttling AI infrastructure development:
Gap 1: Peak demand is spiking as base load generation capacity contracts — 104 GW of firm generation retiring by 2030, replaced by only 22 GW of firm capacity.
Gap 2: Supply chain disruptions are complicating project plans — 65% of executives cite this as a primary challenge.
Gap 3: Long and growing grid build-out timelines — there's currently a seven-year wait on some requests for grid connection.
Gap 4: Cyber and power security are growing concerns — 64% cite them as a top concern.
Gap 5: The permitting process can be long and unpredictable — NEPA reviews, local opposition, multi-year environmental studies.
Gap 6: The industry needs more skilled workers — 63% cite shortage as a top challenge.
Gap 7: Limited pipeline capacity stymies natural gas delivery — constraining even gas-fired backup generation.
The survey's headline finding: 72% of respondents view power and grid capacity constraints as "very or extremely challenging" for data center infrastructure buildout.
As Deloitte concluded: "America's advanced economy relies on 50-year-old infrastructure that cannot meet the increasing electricity demands driven by AI."
The Inescapable Conclusion
Any single constraint discussed in this chapter would be a serious obstacle.
Together, they form a wall that legacy infrastructure cannot climb.
Consider the math:
• Transformers: 4+ year lead times, 30% supply deficit, 80% imported, backdoor security risks in Chinese units
• Gas turbines: 6-7 year lead times, record backlogs, OEM reluctance to expand
• Nuclear SMRs: Mid-2030s at earliest, flagship project canceled, regulatory labyrinth
• Labor: 500,000 electrician shortage, workforce shrinking as demand grows
• Grid interconnection: 5-7 year waits (detailed in Chapter 6)
• Community opposition: $64.4 billion in projects blocked or delayed (detailed in Chapter 7)
Now set these constraints against the demand:
• AI data center demand projected: 123 GW by 2035
• Current total data center demand: 33 GW (2024)
• Net new capacity needed: 90 GW in ten years (very conservative estimate)
The numbers don't reconcile. There is no scenario in which the legacy system supply chain delivers 90 GW of new capacity in a decade when every critical component faces multi-year bottlenecks.
For want of a nail, the kingdom was lost. For want of a transformer, America's AI leadership may be lost — unless we build around the chokepoints.
The Path Forward
The rational conclusion is the one I've spent over six years developing and proving:
The electric grid has passed its expiration date for datacenter loads. New systems must be added immediately. Large electricity customers are better suited to remain off-grid entirely — linked directly to power plants built specifically for their needs.
Off-grid Solar Computing Clusters bypass every bottleneck described in this chapter:
• No transformer constraints — built with standard, available equipment at scales that don't require custom mega-transformers
• No turbine backlogs — solar panels and batteries ship in months, not years
• No nuclear delays — operational now, not in 2035
• No grid interconnection queue — completely off-grid by design
• No electrician shortage at scale — modular, standardized deployment reduces specialized labor requirements
• No community opposition — remote, uninhabited locations where no one objects
• No Chinese backdoors — domestic, secure supply chain for all critical components
I've received numerous patents on this technology and built commercial-scale systems to prove it works. My first commercial-scale prototype went into service over three years ago. It has demonstrated sufficient uptime for modern computing workloads while remaining completely independent of the grid.
The supply chain constraints facing legacy data centers will persist well into the mid-2030s. Every year of waiting is hundreds of $billions in lost revenue, ceded market position, and eroded technological advantage. Those who design around the chokepoints will capture the defining opportunity of the AI era.
Learn more about my work in the next chapters, and at www.639solar.com.
Sources:
For Want of a Nail: The Crash of Emery Worldwide Airlines Flight 17, Admiral Cloudberg
A Look Through History's Most Disastrous Supply Chain Disruptions, Richard Teuchler
Transformer Supply Bottleneck Threatens Power System Stability as Load Grows, UtilityDive
Transformer Troubles: Manufacturing and Policy Constraints Hit US Transformer Supply, WoodMakenzie
Small Modular Reactors: A Realist Approach to the Future of Nuclear Power, ITIF
Addressing the Critical Shortage of Power Transformers to Ensure Reliability of The US Grid, The National Infrastructure Advisory Council
What are Transformers and How They Work, Maddox
The Transformer Crisis: An Industry on the Brink, Power Magazine
Can US Infrastructure Keep Up With the AI Economy?, Deloitte
Turbine Trouble: Why Power Projects are Facing Delays, CFC Solutions
Understanding the Electric Transformer Shortage, Latitude Media
The Country Needs More Electricity and More Electricians, Fox Business
Solving the Electrician Shortage, Forbes
