AI x Energy · Entry 4 of 5

AI Moves at the Speed of Steel

The ceasefire moved in days, oil moved in hours, and hyperscaler money moved in commitments. The physical system barely moved at all. Turbines, transformers, LNG trains, and grid connections were already the binding constraint; the blockade and the $630B hyperscaler pledge simply made that constraint visible to everyone at once.

Apr 19, 2026 · 10 min read

The Story

The Calendar Was Always Physical

Episode 2 left us watching the ceasefire. That was the wrong variable.

The ceasefire moved in days. Oil moved in hours. Political commitments moved in press releases. The machines did not move. Gas turbines still take seven years. Transformers still take three to five. LNG trains still take two to four. ERCOT still cannot absorb 410 GW of large-load requests on any timetable that matters.

The Islamabad talks matter because they proved the point. The diplomatic layer can flip from ceasefire to interdiction inside a week. The physical layer cannot. April 2026 is not the moment the AI-energy story became physical; it was always physical. This is the moment the abstraction failed and the bottleneck became impossible for any participant to ignore.

The headline was the blockade. The lesson is that the backlog had been the story all along.

What Made It Visible

Diplomacy Moved Fast. Infrastructure Did Not.

The Islamabad talks hosted by PM Sharif collapsed after 21 hours of negotiation on April 11-12. VP JD Vance, Steve Witkoff, and Jared Kushner left without a deal; the US and Iran agreed on most of a 10-point framework but could not close on uranium enrichment, the highly enriched stockpile, allied militant groups, or an un-tolled reopening of Hormuz.

A US naval interdiction of Iranian-linked vessels began April 13. Hormuz cycled open and closed twice over the following five days as Tehran and Washington traded statements. Brent traded a roughly $15 intraday range through mid-April without picking a direction, which is the market’s way of saying it no longer expects a clean resolution.

The blockade created immediate fiscal pressure on Iran and immediate price volatility for oil. For the AI-energy story, its more important effect was to raise the value of domestic, dispatchable generation that does not depend on Middle Eastern transit. Treat this section as context, not protagonist: the diplomatic layer is the stress test that revealed where the system was already weak.

(NPR, Time, CNBC blockade, Washington Post, CNBC oil)

The Fuel Layer Has Deadlines

The fuel system did not have a Hormuz problem before; it now has one with calendars attached.

On April 16, the IEA warned Europe could run out of jet fuel in as little as six weeks. Italian airport rationing has expanded to at least 8 airports, disrupting flights to 9 countries. Ireland joined Spain, Germany, Italy, France, and the Netherlands in coordinated emergency reserve releases. Airlines warned of “severe cuts starting in May or June.”

TTF gas prices cooled to €41.25/MWh from the Episode 2 spike of €58.42/MWh, helped by warmer weather and Asian LNG imports running at 6-year lows. European gas storage entered April at 28% (vs 35% a year ago), and the summer refill season is now the most important energy policy variable on the continent. Rebuilding stocks from 28% to normal 90% targets before October, with Qatar partially offline and Hormuz unreliable, is not straightforward.

Qatar mobilized workers to restart Ras Laffan after the April 8 ceasefire window. The full picture: 2 of 14 LNG trains destroyed, one of two GTL facilities damaged, 17% of Qatari export capacity offline. North site restarts within a month. South site not until end of summer. Trains 4 and 6 rebuild: 3-5 years, constrained by only 3 manufacturers worldwide of the required turbomachinery, with 2-4 year equipment lead times. QatarEnergy declared long-term force majeure on contracts with China, South Korea, Italy, and Belgium.

The geographic edge of the shock is sharper in Asia than in Europe. India has lost ~3 million bbl/day of Hormuz-transit crude. LPG (60% of Indian cooking fuel, mostly via Hormuz) is in short supply. Indian refiners are pivoting to Russian Urals; on April 11, the US waiver allowing countries to buy Russian crude expired, coincident with the interdiction start. Japan imports >90% of its crude from the Middle East and authorized older coal-fired plants to run for a 12-month emergency period beginning April 2026, a quiet acknowledgment that energy security beats climate targets during acute crises. China, India, Japan, and South Korea together take 75% of Middle East oil exports and 59% of Middle East LNG exports. The Hormuz disruption is an Asian problem before it is a European one.

(CNBC, The National, Bloomberg Qatar, CNBC India, 19FortyFive)

The Money Already Moved

The capital response is the fastest-moving layer of all. The Big Four hyperscalers are planning up to $630 billion in 2026 capex, up 62% from 2025’s $388B, roughly 4x the entire US electric utility industry’s 2024 investment.

  • Amazon: $200B (from $125B)
  • Google: $175-185B (from $91B)
  • Meta: $115-135B (from $72B)
  • Microsoft: $110-120B (from $90B)

On March 4, seven hyperscalers (Amazon, Google, Meta, Microsoft, OpenAI, Oracle, xAI) signed a White House-brokered agreement committing to build, procure, or fund new generation sufficient for their data center demand and to pay for all grid infrastructure upgrades required to connect them. New load, new generation, new wires, on the hyperscaler’s dime.

The pledge retroactively explains the Microsoft-Chevron Pecos deal (Episode 2), Meta’s 6.6 GW Vistra/Oklo/TerraPower nuclear package (March 2026), Google’s de-facto Goodnight partnership, and Amazon’s X-energy commitment. It also resolves the main political objection to hyperscaler siting: ratepayers are no longer subsidizing AI.

ERCOT’s queue is the physical rendering of all this money looking for an outlet. The 225 requests we tracked in Episode 2 is now 410 GW of large-load interconnection requests, 87% from data centers, against an all-time ERCOT peak demand of ~87 GW. The PUCT voted March 12 to publish the draft SB-6 interconnection rule (16 TAC §25.194), imposing substantial financial obligations and disclosure requirements on loads ≥75 MW. ERCOT is pushing to approve its “Batch Zero” admission criteria by June 2026, with the first batch study beginning late summer. By design or consequence, SB-6 is a behind-the-meter accelerator: every friction added to grid connection is another reason to self-generate on-site.

The point: the money and the policy are no longer the binding variables. They have already moved.

(POWER Magazine, CreditSights, Data Center Richness, Latitude Media, Perkins Coie, Inside Climate News)

The Equipment Layer Is The Bottleneck

This is where the story actually lives.

Gas turbines. GE Vernova reports Q1 earnings on April 22. CEO Scott Strazik has pre-guided Q1 gas contracts of 12-24 GW (vs 8 GW Q1 2025). The year-end 2025 gas turbine backlog hit 83 GW (up from 62 GW one quarter earlier); company target is 100 GW by year-end 2026. Total firm-wide backlog: a record $150.2B with $85B services component. Multi-year target: $200B. Microsoft-Chevron-Engine No. 1 placed a specific order for 7 large GE Vernova natural gas turbines as part of the $7B Pecos deal.

Siemens Energy booked 100+ gas turbines in Q1 FY26, half its 2025 full-year volume in a single quarter. Gas Services order intake: €17.6B ($21B), an all-time record. Regional split 40% US / 35% Europe / 15% Middle East+China. Wait times up to 7 years. Siemens is resuming gas turbine manufacturing in Charlotte, NC as part of its $1B US expansion.

Mitsubishi Power booked 6 M501JAC hydrogen-ready turbines for Saudi Arabia totaling 3.6 GW, and announced that all new European projects are hydrogen-ready by default. Mitsubishi is raising production capacity 30%.

Three OEMs. Seven-year wait times. The order book is closed for the rest of the decade for everyone who did not commit early.

Transformers are the harder constraint. Analyst reports now suggest more than half of US data centers planned for 2026 may be delayed or canceled, not for lack of turbines but because of transformer shortages. Large power transformer lead times have stretched from 24-30 months to 3-5 years.

Hitachi Energy CEO Andreas Schierenbeck said the transformer sector is “overwhelmed” and committed $6B / 15,000 hires over 3 years. A $457M facility in South Boston, Virginia will become the largest US LPT plant by 2028; a $106M Alamo, Tennessee expansion targets critical components. Every behind-the-meter gas plant, every data center, every grid upgrade needs high-capacity transformers. A 3-5 year lead time means transformers ordered today deliver in 2029-2031, the same window as the sold-out gas turbine backlog. Transformer OEMs (Hitachi Energy, Eaton, ABB, Siemens Energy transformer division, Hyundai Electric) now have pricing power on par with turbine OEMs.

The equipment layer is not responding to news. It is responding to physical manufacturing capacity, and that capacity is the variable that governs everything else in this story.

(Yahoo Finance, Utility Dive, Bloomberg Siemens, S&P Global Mitsubishi, DCD, Bloomberg DC equipment, Sandstone Group)

The Side Signals Point The Same Way

X-energy filed for IPO on Nasdaq (ticker: XE), seeking up to $814M. Orderbook >11 GW (~144 Xe-100 SMRs). Amazon has a direct investment plus 5+ GW deployment option by 2039. First reference project: Cascade Advanced Energy Facility in Washington State.

Venture Global CP2 Phase 2 FID - the first US LNG export project sanctioned in 2026. 20 MTPA, 36 modular trains, Cameron Parish LA. Combined with the 13% Plaquemines expansion, Venture Global is positioning to challenge Cheniere as the largest US LNG exporter. (NGI)

Google’s Goodnight emission numbers disclosed in Crusoe’s permit: up to 4.5M tons CO2/year from the 933 MW gas plant - equivalent to San Francisco’s entire emissions. Google: “We don’t have a contract in place for the plant in Texas.” Satellite imagery shows construction is underway. (Yale E360)

OPEC+ approved 206k bbl/day increase for May, same as April. Symbolic against Hormuz disruption. Gulf producers can raise quota; they cannot physically ship through a closed strait. (Rigzone)

Jones Act suspended for 60 days (mid-March). Enables foreign-flagged vessels to move oil, LNG, fertilizer, coal between US ports. Significant because zero US-flagged LNG carriers exist - the suspension lets Puerto Rico finally import from Gulf Coast terminals. Expires mid-May. (PBS)

SPEED Act still has no Senate draft. The ~8-week window we flagged in Episode 2 (through early May) has ~2 weeks left. Senate Democrats engaged in early March but no text has emerged. Transmission and Trump’s wind/solar permit revocation authority remain unresolved. The chance of passage before midterms nudges back down. (Arnold & Porter)

KIT’s compressorless hydrogen turbine demos at Hannover Messe April 20-24 (Hall 11, Stand B 06). 303-second runtime extended past 5 minutes. Lab-scale but structurally disruptive if it scales. (KIT)

Conclusions

The Hormuz disruption did not change the thesis; it removed the option of ignoring it. Middle Eastern oil and LNG can no longer be priced as reliable inputs for multi-year infrastructure planning, and every downstream decision now sits on top of that assumption.

Qatar’s 3-5 year Ras Laffan rebuild is the most durable consequence of this cycle. Ceasefires can unwind and interdictions can be lifted, but destroyed LNG trains rebuild at the pace of turbomachinery manufacturing, not diplomacy. That clears the field for US LNG exporters through roughly 2029-2031.

The $630B hyperscaler capex + White House pledge turns the “who pays for transmission” debate from blocker to non-issue wherever hyperscalers show up. Capital and policy are no longer the constraint. Combined with Qatar’s loss and European demand, US gas infrastructure capex has an unprecedented multi-year demand signal pointing at the same equipment makers everyone else is queuing for.

ERCOT’s 410 GW queue + SB-6 is the physical expression of money meeting calendar. Texas cannot connect 410 GW of grid-connected load in five years, so hyperscalers go behind-the-meter, which is exactly what the policy regime quietly rewards.

The transformer shortage is the under-priced constraint. Gas turbines are famously sold out through 2029-2031 with 7-year Siemens wait times. Transformers have 3-5 year lead times and fewer global manufacturers. If transformer supply is the actual binding constraint, a meaningful share of the 2026-2027 data center pipeline slips regardless of what SPEED or hyperscaler capex does.

India’s waiver expiration coinciding with the interdiction makes a US-India bilateral energy package likely in the next 60-90 days, with US LNG and crude exchanged for geopolitical alignment.

Japan authorizing emergency coal and Europe rationing jet fuel converge on the same political lesson: in acute crises, energy security beats climate targets. That is tailwind for gas and nuclear, and headwind for offshore wind or solar-as-primary-capacity arguments.


What We Are Watching

  • Whether GE Vernova and Siemens confirm that turbine demand is still accelerating.
  • Whether transformer lead times become the visible reason data center projects slip.
  • Whether Qatar’s rebuild timeline pushes more LNG buyers toward US supply.
  • Whether ERCOT’s Batch Zero process turns into a behind-the-meter accelerator.
  • Whether hyperscaler capex keeps moving faster than equipment manufacturing.

The Five-Year Note

The money is committed. The politics are aligned. The demand is structural. The only remaining constraint is physics.

Turbines take 7 years. Transformers take 3-5 years. LNG trains take 2-4 years. The blockade might end. A ceasefire might return. Oil might fall for a week. None of that pulls a transformer forward by three years or creates an unused turbine slot. Pecos does not come online until 2027 and Goodnight not until 2028, and there is no way to pull those dates forward.

When rereading this later, the thing to remember is not whether the blockade lasted. It is that by April 2026 the physical bottleneck that had always governed the story became visible to everyone at once. The race between physical infrastructure delivery and cumulative AI + energy demand is the story of the next 48 months. Tracking the order book, the backlog, and the delivery calendar is the work now, because those are the decisive variables.

Field Notes

The source-layer research that backs this episode.

  1. 021 Energy Services: SLB & Baker Hughes Q1 Previews
  2. 022 ERCOT Queue Hits 410 GW of Large Load Requests
  3. 023 Europe Jet Fuel: IEA Warns Six Weeks Left
  4. 024 GE Vernova Q1 Preview: 83 GW Backlog, Earnings April 22
  5. 025 Google-Crusoe Goodnight: Permit Emissions Disclosed
  6. 026 Hyperscaler $630B CapEx and White House Power Pledge
  7. 027 India and Japan: The Most Exposed Major Economies
  8. 028 Islamabad Talks Collapse and Blockade Begins
  9. 029 Jones Act 60-Day Suspension
  10. 030 KIT Compressorless Hydrogen Turbine - Hannover Messe Demo
  11. 031 Mitsubishi Power: 3.6 GW Saudi JAC Order
  12. 032 OPEC+ May 206k Hike & US LNG Expansion
  13. 033 Qatar Ras Laffan LNG Restart Timeline
  14. 034 Siemens Energy Books Record Gas Turbine Orders
  15. 035 SPEED Act: Window Closing, Still No Senate Draft
  16. 036 Transformer Shortage: Half of 2026 Data Center Pipeline at Risk
  17. 037 US Naval Blockade of the Strait of Hormuz
  18. 038 X-energy Files for Nuclear IPO
← AI x Energy
AI x Energy · Entry 4 of 5

AI Moves at the Speed of Steel

The ceasefire moved in days, oil moved in hours, and hyperscaler money moved in commitments. The physical system barely moved at all. Turbines, transformers, LNG trains, and grid connections were already the binding constraint; the blockade and the $630B hyperscaler pledge simply made that constraint visible to everyone at once.

Apr 19, 2026 · 10 min read

The Story

The Calendar Was Always Physical

Episode 2 left us watching the ceasefire. That was the wrong variable.

The ceasefire moved in days. Oil moved in hours. Political commitments moved in press releases. The machines did not move. Gas turbines still take seven years. Transformers still take three to five. LNG trains still take two to four. ERCOT still cannot absorb 410 GW of large-load requests on any timetable that matters.

The Islamabad talks matter because they proved the point. The diplomatic layer can flip from ceasefire to interdiction inside a week. The physical layer cannot. April 2026 is not the moment the AI-energy story became physical; it was always physical. This is the moment the abstraction failed and the bottleneck became impossible for any participant to ignore.

The headline was the blockade. The lesson is that the backlog had been the story all along.

What Made It Visible

Diplomacy Moved Fast. Infrastructure Did Not.

The Islamabad talks hosted by PM Sharif collapsed after 21 hours of negotiation on April 11-12. VP JD Vance, Steve Witkoff, and Jared Kushner left without a deal; the US and Iran agreed on most of a 10-point framework but could not close on uranium enrichment, the highly enriched stockpile, allied militant groups, or an un-tolled reopening of Hormuz.

A US naval interdiction of Iranian-linked vessels began April 13. Hormuz cycled open and closed twice over the following five days as Tehran and Washington traded statements. Brent traded a roughly $15 intraday range through mid-April without picking a direction, which is the market’s way of saying it no longer expects a clean resolution.

The blockade created immediate fiscal pressure on Iran and immediate price volatility for oil. For the AI-energy story, its more important effect was to raise the value of domestic, dispatchable generation that does not depend on Middle Eastern transit. Treat this section as context, not protagonist: the diplomatic layer is the stress test that revealed where the system was already weak.

(NPR, Time, CNBC blockade, Washington Post, CNBC oil)

The Fuel Layer Has Deadlines

The fuel system did not have a Hormuz problem before; it now has one with calendars attached.

On April 16, the IEA warned Europe could run out of jet fuel in as little as six weeks. Italian airport rationing has expanded to at least 8 airports, disrupting flights to 9 countries. Ireland joined Spain, Germany, Italy, France, and the Netherlands in coordinated emergency reserve releases. Airlines warned of “severe cuts starting in May or June.”

TTF gas prices cooled to €41.25/MWh from the Episode 2 spike of €58.42/MWh, helped by warmer weather and Asian LNG imports running at 6-year lows. European gas storage entered April at 28% (vs 35% a year ago), and the summer refill season is now the most important energy policy variable on the continent. Rebuilding stocks from 28% to normal 90% targets before October, with Qatar partially offline and Hormuz unreliable, is not straightforward.

Qatar mobilized workers to restart Ras Laffan after the April 8 ceasefire window. The full picture: 2 of 14 LNG trains destroyed, one of two GTL facilities damaged, 17% of Qatari export capacity offline. North site restarts within a month. South site not until end of summer. Trains 4 and 6 rebuild: 3-5 years, constrained by only 3 manufacturers worldwide of the required turbomachinery, with 2-4 year equipment lead times. QatarEnergy declared long-term force majeure on contracts with China, South Korea, Italy, and Belgium.

The geographic edge of the shock is sharper in Asia than in Europe. India has lost ~3 million bbl/day of Hormuz-transit crude. LPG (60% of Indian cooking fuel, mostly via Hormuz) is in short supply. Indian refiners are pivoting to Russian Urals; on April 11, the US waiver allowing countries to buy Russian crude expired, coincident with the interdiction start. Japan imports >90% of its crude from the Middle East and authorized older coal-fired plants to run for a 12-month emergency period beginning April 2026, a quiet acknowledgment that energy security beats climate targets during acute crises. China, India, Japan, and South Korea together take 75% of Middle East oil exports and 59% of Middle East LNG exports. The Hormuz disruption is an Asian problem before it is a European one.

(CNBC, The National, Bloomberg Qatar, CNBC India, 19FortyFive)

The Money Already Moved

The capital response is the fastest-moving layer of all. The Big Four hyperscalers are planning up to $630 billion in 2026 capex, up 62% from 2025’s $388B, roughly 4x the entire US electric utility industry’s 2024 investment.

On March 4, seven hyperscalers (Amazon, Google, Meta, Microsoft, OpenAI, Oracle, xAI) signed a White House-brokered agreement committing to build, procure, or fund new generation sufficient for their data center demand and to pay for all grid infrastructure upgrades required to connect them. New load, new generation, new wires, on the hyperscaler’s dime.

The pledge retroactively explains the Microsoft-Chevron Pecos deal (Episode 2), Meta’s 6.6 GW Vistra/Oklo/TerraPower nuclear package (March 2026), Google’s de-facto Goodnight partnership, and Amazon’s X-energy commitment. It also resolves the main political objection to hyperscaler siting: ratepayers are no longer subsidizing AI.

ERCOT’s queue is the physical rendering of all this money looking for an outlet. The 225 requests we tracked in Episode 2 is now 410 GW of large-load interconnection requests, 87% from data centers, against an all-time ERCOT peak demand of ~87 GW. The PUCT voted March 12 to publish the draft SB-6 interconnection rule (16 TAC §25.194), imposing substantial financial obligations and disclosure requirements on loads ≥75 MW. ERCOT is pushing to approve its “Batch Zero” admission criteria by June 2026, with the first batch study beginning late summer. By design or consequence, SB-6 is a behind-the-meter accelerator: every friction added to grid connection is another reason to self-generate on-site.

The point: the money and the policy are no longer the binding variables. They have already moved.

(POWER Magazine, CreditSights, Data Center Richness, Latitude Media, Perkins Coie, Inside Climate News)

The Equipment Layer Is The Bottleneck

This is where the story actually lives.

Gas turbines. GE Vernova reports Q1 earnings on April 22. CEO Scott Strazik has pre-guided Q1 gas contracts of 12-24 GW (vs 8 GW Q1 2025). The year-end 2025 gas turbine backlog hit 83 GW (up from 62 GW one quarter earlier); company target is 100 GW by year-end 2026. Total firm-wide backlog: a record $150.2B with $85B services component. Multi-year target: $200B. Microsoft-Chevron-Engine No. 1 placed a specific order for 7 large GE Vernova natural gas turbines as part of the $7B Pecos deal.

Siemens Energy booked 100+ gas turbines in Q1 FY26, half its 2025 full-year volume in a single quarter. Gas Services order intake: €17.6B ($21B), an all-time record. Regional split 40% US / 35% Europe / 15% Middle East+China. Wait times up to 7 years. Siemens is resuming gas turbine manufacturing in Charlotte, NC as part of its $1B US expansion.

Mitsubishi Power booked 6 M501JAC hydrogen-ready turbines for Saudi Arabia totaling 3.6 GW, and announced that all new European projects are hydrogen-ready by default. Mitsubishi is raising production capacity 30%.

Three OEMs. Seven-year wait times. The order book is closed for the rest of the decade for everyone who did not commit early.

Transformers are the harder constraint. Analyst reports now suggest more than half of US data centers planned for 2026 may be delayed or canceled, not for lack of turbines but because of transformer shortages. Large power transformer lead times have stretched from 24-30 months to 3-5 years.

Hitachi Energy CEO Andreas Schierenbeck said the transformer sector is “overwhelmed” and committed $6B / 15,000 hires over 3 years. A $457M facility in South Boston, Virginia will become the largest US LPT plant by 2028; a $106M Alamo, Tennessee expansion targets critical components. Every behind-the-meter gas plant, every data center, every grid upgrade needs high-capacity transformers. A 3-5 year lead time means transformers ordered today deliver in 2029-2031, the same window as the sold-out gas turbine backlog. Transformer OEMs (Hitachi Energy, Eaton, ABB, Siemens Energy transformer division, Hyundai Electric) now have pricing power on par with turbine OEMs.

The equipment layer is not responding to news. It is responding to physical manufacturing capacity, and that capacity is the variable that governs everything else in this story.

(Yahoo Finance, Utility Dive, Bloomberg Siemens, S&P Global Mitsubishi, DCD, Bloomberg DC equipment, Sandstone Group)

The Side Signals Point The Same Way

X-energy filed for IPO on Nasdaq (ticker: XE), seeking up to $814M. Orderbook >11 GW (~144 Xe-100 SMRs). Amazon has a direct investment plus 5+ GW deployment option by 2039. First reference project: Cascade Advanced Energy Facility in Washington State.

Venture Global CP2 Phase 2 FID - the first US LNG export project sanctioned in 2026. 20 MTPA, 36 modular trains, Cameron Parish LA. Combined with the 13% Plaquemines expansion, Venture Global is positioning to challenge Cheniere as the largest US LNG exporter. (NGI)

Google’s Goodnight emission numbers disclosed in Crusoe’s permit: up to 4.5M tons CO2/year from the 933 MW gas plant - equivalent to San Francisco’s entire emissions. Google: “We don’t have a contract in place for the plant in Texas.” Satellite imagery shows construction is underway. (Yale E360)

OPEC+ approved 206k bbl/day increase for May, same as April. Symbolic against Hormuz disruption. Gulf producers can raise quota; they cannot physically ship through a closed strait. (Rigzone)

Jones Act suspended for 60 days (mid-March). Enables foreign-flagged vessels to move oil, LNG, fertilizer, coal between US ports. Significant because zero US-flagged LNG carriers exist - the suspension lets Puerto Rico finally import from Gulf Coast terminals. Expires mid-May. (PBS)

SPEED Act still has no Senate draft. The ~8-week window we flagged in Episode 2 (through early May) has ~2 weeks left. Senate Democrats engaged in early March but no text has emerged. Transmission and Trump’s wind/solar permit revocation authority remain unresolved. The chance of passage before midterms nudges back down. (Arnold & Porter)

KIT’s compressorless hydrogen turbine demos at Hannover Messe April 20-24 (Hall 11, Stand B 06). 303-second runtime extended past 5 minutes. Lab-scale but structurally disruptive if it scales. (KIT)

Conclusions

The Hormuz disruption did not change the thesis; it removed the option of ignoring it. Middle Eastern oil and LNG can no longer be priced as reliable inputs for multi-year infrastructure planning, and every downstream decision now sits on top of that assumption.

Qatar’s 3-5 year Ras Laffan rebuild is the most durable consequence of this cycle. Ceasefires can unwind and interdictions can be lifted, but destroyed LNG trains rebuild at the pace of turbomachinery manufacturing, not diplomacy. That clears the field for US LNG exporters through roughly 2029-2031.

The $630B hyperscaler capex + White House pledge turns the “who pays for transmission” debate from blocker to non-issue wherever hyperscalers show up. Capital and policy are no longer the constraint. Combined with Qatar’s loss and European demand, US gas infrastructure capex has an unprecedented multi-year demand signal pointing at the same equipment makers everyone else is queuing for.

ERCOT’s 410 GW queue + SB-6 is the physical expression of money meeting calendar. Texas cannot connect 410 GW of grid-connected load in five years, so hyperscalers go behind-the-meter, which is exactly what the policy regime quietly rewards.

The transformer shortage is the under-priced constraint. Gas turbines are famously sold out through 2029-2031 with 7-year Siemens wait times. Transformers have 3-5 year lead times and fewer global manufacturers. If transformer supply is the actual binding constraint, a meaningful share of the 2026-2027 data center pipeline slips regardless of what SPEED or hyperscaler capex does.

India’s waiver expiration coinciding with the interdiction makes a US-India bilateral energy package likely in the next 60-90 days, with US LNG and crude exchanged for geopolitical alignment.

Japan authorizing emergency coal and Europe rationing jet fuel converge on the same political lesson: in acute crises, energy security beats climate targets. That is tailwind for gas and nuclear, and headwind for offshore wind or solar-as-primary-capacity arguments.


What We Are Watching

The Five-Year Note

The money is committed. The politics are aligned. The demand is structural. The only remaining constraint is physics.

Turbines take 7 years. Transformers take 3-5 years. LNG trains take 2-4 years. The blockade might end. A ceasefire might return. Oil might fall for a week. None of that pulls a transformer forward by three years or creates an unused turbine slot. Pecos does not come online until 2027 and Goodnight not until 2028, and there is no way to pull those dates forward.

When rereading this later, the thing to remember is not whether the blockade lasted. It is that by April 2026 the physical bottleneck that had always governed the story became visible to everyone at once. The race between physical infrastructure delivery and cumulative AI + energy demand is the story of the next 48 months. Tracking the order book, the backlog, and the delivery calendar is the work now, because those are the decisive variables.

Map

The ideas this entry touches, and where they show up elsewhere.