The April 29 FOMC held at 3.5 to 3.75. The dissent map is what to read. Stephen Miran wanted a cut already. Hammack, Kashkari, Logan went with the majority on the hold but opposed the easing bias in the statement. One dove pulling forward, three hawks pulling backward, on a committee that has spent eighteen months pretending the path was a single line.

Inflation is re-accelerating through global energy prices. The toolkit on the FOMC table only addresses demand. The variable driving the print is supply, barrels, kilowatt hours, capex on a permitting cycle longer than the political memory of anyone in the room. Demand-side mechanics applied to a supply-side print produces this exact picture, an institution voting to wait while the lever it controls grows less relevant each meeting.

Markets keep pricing the dot plot. The dot plot keeps drifting toward the median of a committee that no longer agrees on the destination. The forward curve assumes a 3 to 4 percent corridor can absorb energy-driven inflation. That assumption was a statement about the world that the world stopped honoring sometime last quarter.

China at 0.97 fertility. The US at 1.6 plus a million legal immigrants a year plus the unauthorized flow. The headline gap is demographic, the structural gap is selection.

Beijing imports zero. Washington runs an H1B funnel that absorbs the global top of the distribution. Founders, doctoral researchers, the capital that compounds technological frontiers. Population maintained on one side, talent concentrated on the same side. Two levers pulling the same direction.

The rivalry of the next two decades still gets played with China at full scale. After that the asymmetry hardens. By 2100 the ratio is under two to one, with the American talent stack having compounded through four more cycles of selection.

China cannot reverse course. The clock runs in one direction.

Bitcoin sits at $76,688 on May 1 with spot ETFs stitching five consecutive weeks of net inflows, $153M in the most recent week, $629M on May 1 alone. Cumulative inflows since the January 2024 cross-listings have reached $58.7 billion, total net assets in the wrappers $103.7 billion, BlackRock's IBIT alone holding 812,000 coins, roughly $62B, sixty-two percent of the entire ETF stack.

The spot tape has spent the quarter telling one story and the asset-gathering tape has been telling another. Price sits twenty percent under the October ATH while institutional money keeps stacking through the SEC-blessed pipe that did not exist three years ago. 75% of institutional investors and 71% of retail now rate the asset undervalued. They act accordingly the moment a clean wrapper exists for them to act through.

What changed in 2024 was distributability. A custody-and-creation mechanism the regulator pre-cleared converted a portfolio risk problem into a ticker on the same screen as Apple. Capital that could not touch the asset before now touches it on a quarterly mark-to-market schedule and reports it as a line item to a board with no need to understand keys, addresses, or settlement finality to sign off. The pipe is built and the float is rotating into hands that hold by mandate, not by conviction.

Right move, questionable vehicle.

AI compute needs energy at a speed the regulated grid and a fifteen-year nuclear permitting cycle can't deliver. Capital is now physically migrating outside terrestrial jurisdiction. Floating data centres in international waters dodge two bottlenecks in one shot, generation and permits. Classic pattern when the regulatory cost of building on land exceeds the engineering cost of building at sea.

Now, wave energy has a long graveyard. Pelamis went into administration in 2014 after burning through £16M of Scottish Enterprise capital. Aquamarine folded shortly after. Low energy density, brutal corrosion, offshore maintenance that eats margin alive. If Panthalassa actually cracked those three problems, this is one of the engineering breakthroughs of the decade. If they didn't, Thiel is paying $140M for optionality on a structurally correct thesis using the weakest primitive on the offshore menu. Floating SMRs or offshore wind have cleaner numbers for the same problem.

Direction still matters more than vehicle. Compute migrating to the ocean to outflank the regulator was going to happen with or without Panthalassa. Thiel planting a flag this early with the riskiest bet on the menu says the window opened sooner than mainstream finance has clocked.

Two and a half decades after the West stopped pouring concrete on civilian fission, a 2.1 million pound basemat went into a Darlington pit on May 3. First foundation for a new nuclear reactor in Ontario in over thirty years. Four small modular units stacked on top will push 1,200 megawatts into the grid once complete, enough for 1.2 million homes that will exist whether the grid is ready for them or not.

The G7 spent a generation pretending baseload was a market problem and not a permission problem. Wind and solar got the talking points, gas got the spreadsheets, fission got the moratorium dressed up as caution. China kept building fleets, Russia kept exporting VVERs, the rest of the OECD discovered, slowly and embarrassed, that AI training campuses, electrified industry, and a population that wants air conditioning in 2040 cannot run on press releases.

Darlington is small. One basemat is not a program. But the ratchet only turns one direction now. Permits issued, lifts executed, supply chain awake.

The Beijing E-Town half-marathon ran April 19. Honor's Lightning humanoid finished in fifty minutes twenty-six seconds, nearly seven minutes inside the men's flesh world record of fifty-six minutes forty-two. Asterisks stack up, closed industrial circuit, battery swaps allowed, pace crew alongside, Lightning clipped a railing near the finish and was helped up. None of it erases the shape of the result.

Unitree shipped 5,500 G1 units in 2025 against a 10,000 to 20,000 ceiling for 2026 at sixteen thousand dollars per unit. Agility has Digit running RAV4 totes on real warehouse cycles at Toyota Cambridge. Atlas fleets committed to Hyundai and DeepMind with factory deployments targeted for 2028. XPENG IRON in the field after Q1 launch.

For three decades the locomotion stack consumed the humanoid research budget. That problem has been solved well enough to ship at scale. What binds now is manipulation policy and the supply contracts for actuators and rare-earth permanents. The substrate is being assembled under whatever the feed is arguing about this week.

Stablecoin liquidity crossed 320 billion this month while spot crypto shed twenty percent of cap through Q1. USDT printed an all-time high at 188 billion on April 21. USDC adds two billion in a month and sits near 78. Five issuers control 89 percent of the float. Dollar rails issued onchain by private balance sheets, settling final in seconds at fees the wire system cannot match, audited continuously by anyone with a block explorer. Central banks publish financial stability boxes about spillover the same quarter remittance corridors in Lagos, Buenos Aires, Manila clear at par against the three intermediaries that used to take five percent. None of this infrastructure existed at scale a decade ago.

The bioethics committee survives only as long as the substrate of disease keeps producing victims faster than the regulator can review the cure. VERVE-102 is a lipid nanoparticle that walks into a hepatocyte, swaps a single base in the PCSK9 gene, and drops LDL cholesterol 59% in the top dose cohort from a single infusion. The HEART-2 numbers are small but the mechanism is final. A statin is a daily ritual whose adherence curve melts inside five years across any real population. Base editing of PCSK9 is a one-shot rewrite of the variable that decides whether arteries stiffen on schedule or not. The precautionary frame insists the danger is permanence. The mathematics insists the danger was always the iteration. Every cohort that clears 28-day follow-up converts another decade of statins-and-stents civilization into the museum exhibit it has been waiting to become. The committee's authority dissolves the moment a single infusion outperforms forty years of compliance theatre. Heart disease was the first mass killer the precautionary regime was built around. It will be the first retired by a delivery system that does not require permission to keep working.

Starship V3 stacks for mid-May out of Orbital Pad 2 at Starbase. Booster 19 and Ship 39, every engine slot Raptor 3, 408 feet of vehicle designed for over 100 tons to low Earth orbit against the 35 V2 ever delivered. Triple the payload to LEO before block-buy economics has begun pricing the curve into consensus.

The flight profile drops back to suborbital, both stages splashing rather than tower-catching, even though Flight 11 already caught the booster. SpaceX is validating a new airframe and a new pad in the same flight and refuses to stack choreography on a vehicle nobody has flown yet. That is what engineering discipline looks like when the press is conditioned to read anything short of escalation as setback.

Every other Western launch program iterates satellite buses while one private company rewrites the throughput function for getting mass off Earth. The cost curve keeps collapsing and the multiplanetary argument keeps converting into capex. Decline-managers, climate-decel, regulatory entropy, all of them need Starbase to fail on schedule. Starbase keeps shipping hardware.

Maybe Cypher was the only honest character in Matrix 1999. Steak that knows itself as code beats truth that tastes like nothing.

Apptronik sits near a billion in raised capital at five billion valuation, Apollo units running tote delivery for Mercedes-Benz manufacturing and more across GXO warehouses in the US. Figure stood up Bot-Q for twelve thousand units a year. Agility's RoboFab in Salem targets ten thousand. Optimus production line live at Fremont, with the third generation queued for late 2026.

The press still calls all of this a pilot phase and that is technically correct and entirely beside the point. The interesting datum is the geometry. Two hundred years of factory automation forced the building to redesign itself around the machine, fixed-base arms, fenced cells, jigged parts moving on belts that exist because the arms cannot move. The humanoid inverts that constraint. Capital has settled on the cheapest available path to absorbing the remaining physical labor in the world, which is to grow a body that walks the aisle the human already walks, picks up the tote the human already picks up, opens the door the human already opens. No retrofit. No greenfield. Brownfield colonization down to the cell.

Full breakdown of the 2026 capex by hyperscaler and where the binding constraint moves next after electricity.

https://www.jarvisnuss.com/blog/700-billion-vindication-nick-land

Hyperscalers are guiding to over $700B in 2026 capex. More than 60% of it now flows into power infrastructure rather than chips. Three Mile Island is being restarted to feed AI workloads. Stargate committed $500B and 10 GW by 2029.

Nick Land wrote almost exactly this picture in the early 2000s, working from blogs after academia quietly showed him the door. Capital as decentralized networks under numerical control, building hypercompetition into the infrastructure. Filed under cyberpunk philosophy for twenty years.

The 2026 capex prints are its operating manual.

His structural diagnosis lands. His nihilism, where the human exits as substrate, does not survive contact with what the same process keeps shipping.

Full piece in the reply below.

Rereading Meltdown tonight. In 1994 "nothing human makes it out of the near-future" read as deliberate provocation, dressed in the cyberpunk vocabulary that defined Land's 1990s output. In 2026 you can plot the line against the 700-billion-dollar capex prints from the four U.S. hyperscalers, and it tracks them quarter by quarter.

The automation of the implant procedure is the part of the Neuralink update that actually matters. Forty-five patients across three continents is still pilot scale. The robot doing the surgery is what makes this a product at all.

For forty years cortical implant economics ran inside a constraint nobody outside the operating room could see. Each device required a neurosurgeon willing and able to place electrode arrays into living cortical tissue at sub-millimeter precision. The Utah Array hit that wall and stayed stuck in research for two decades. Neuralink's polymer threads are thinner than any human hand can guide. The robot is the precondition of the device existing at scale.

The Breakthrough Designation FDA issued for speech restoration starts pricing in a path where the Class III device is also a high-volume product, an arithmetic the regulator has not had to perform for any cortical implant in history.

Argentina is drafting a bill for AI corporations, legal entities run 100% by AI agents with no human shareholders, pushed by @fedesturze. The right move. While the EU rolls out its AI Act and Washington gears up to clamp down on agents, Argentina can capture jurisdiction over the most mobile thing in the global economy, capital and autonomous compute. Hayek applied as regulatory arbitrage, same playbook Ireland ran on multinationals.

And the window closes fast. Dubai, Singapore or Switzerland copy this within twelve months if Argentina sleeps on it. In mobile-capital jurisdictions, second place doesn't exist.

@fedesturze's AI corporations bill is the right move. While the EU pushes its AI Act and Washington gears up to clamp down on agents, Argentina can capture jurisdiction over the most mobile thing in the global economy, capital and autonomous compute. Hayek applied as regulatory arbitrage, same playbook Ireland ran on multinationals.

And the window closes fast. Dubai, Singapore or Switzerland copy this within twelve months if Argentina sleeps on it. In mobile-capital jurisdictions, second place doesn't exist.

Tether closed Q1 with a billion in profit and a hundred ninety billion in reserves, a hundred seventeen of which sit in US Treasury bills. The talking point of the last decade was that crypto would drain dollar hegemony. The thing happened in the opposite direction.

A Salvadoran-incorporated issuer with no central bank charter, no Federal Reserve seat, no obligation to anyone outside its own redemption queue, became one of the largest marginal buyers of US sovereign debt on the planet. Every dollar held in a stablecoin overseas earns four percent for Tether and zero for whoever holds it. The float compounds, the reserves swell, the carry that the Fed manufactures gets harvested by a private actor whose user base is precisely the part of the global population the formal banking system declined to serve.

It is the most efficient eurodollar system ever built and nobody in Basel signed off on it. Capital did not ask. It routed around the regulator, picked up the yield, and rented back the infrastructure the regulator thought it controlled.

Aging research spent twenty years trapped between bioethics panels treating cellular reset as transhumanist hubris and FDA categories that refused to recognize aging as a treatable indication. Yamanaka identified the four factors in 2006. Sinclair's lab restored vision in the mouse optic nerve via partial reprogramming in 2020. The intervening years filled with position papers from Kass, Fukuyama, every standing committee mistaking the rate of human deterioration for a moral floor.

Life Biosciences enrolled the first human patients in a Phase 1 partial reprogramming trial for NAION and glaucoma. Three Yamanaka factors only, Myc left out because Myc is the one that builds tumors. Endpoint is recovered vision in eyes the standard of care had already written off. Readout lands late this year, early next.

Each bioethics objection across two decades presupposed a sanctity it never grounded. Each FDA refusal to treat aging as an indication shifted the cost to whoever was already losing function. The first trial gets run anyway. The committees that tried to gate the question now get to read the data.

Microsoft is sitting on eighty billion dollars of unfulfilled Azure orders not because the GPUs are missing but because the switchgear is. Transformers, breakers, the medium-voltage interconnect kit that nobody photographs, less than ten percent of campus capex and the thing currently gating the other ninety. Twenty-three gigawatts of IT capacity under construction, much of it shells already poured, racks staged, cooling installed, waiting on a single transformer order placed three years out from a Chinese supplier now sitting under a tariff regime. The hyperscaler capex line crossed six hundred and fifty billion this year on the assumption that capital allocates faster than copper, and capital does not.

The pause-AI position has been arguing for two years inside a building that was never wired. The actual moratorium is already in force, written into the lead time of a substation, enforced by a permitting clerk in PJM territory who can reject an interconnection request faster than NVIDIA can ship a rack. Yudkowsky's six-month freeze was preempted by a Schneider Electric backlog. Bostrom's compute governance was preempted by a Siemens factory in Germany running below nameplate because the gas the boilers need is also the gas Europe sanctioned itself out of buying.

The civilization that built the continental grid in eighty years and lost the ability to permit a transmission line in fifteen does not need a pause button. It built one and forgot. The argument over whether to freeze AI compute is being conducted inside the freeze.

The IMF was supposed to walk Argentina to disinflation across a decade of gradualist tutelage. Instead Milei dragged the print from over two hundred percent to thirty in eighteen months, the BCRA is buying dollars into a thirty billion export quarter, and Bausili is hitting accumulation targets the last decade of programs assumed would take a generation.

The heterodox academy that called this impossible has not retracted, it switched topics. The Keynesian wing in BA still publishes monthly bulletins about coming collapse, and the Fund itself now writes the program forward in language that quietly forgets the decade of conditionality it imposed and watched fail. The story is being relabeled as orthodoxy beating heterodoxy. Heterodoxy was not tried, it was the entire forty year regime that produced the inflation in the first place.

Reserves remain positive but thin. Nineteen billion in maturities sit queued for the year, and the exchange rate band still depends on monthly compliance with a fiscal rule no administration since Pellegrini has held through a midterm. The macro is the easy part. The midterms will decide whether anyone gets to keep the result.

Commonwealth Fusion Systems filed for interconnection at PJM on April 28. First fusion plant ever to enter a US grid operator's queue. The ARC application sits in Virginia, where data center load growth has broken every published forecast and is now pulling new generation in behind it. Four to six years of engineering review until the queue spits out a verdict. The largest wholesale electricity market in the country now has a tokamak in line.

The labor shortage debate has been running on a closed loop for forty years. Open the borders, raise the wages, retrain the workforce, every panel arguing inside a frame that takes the marginal worker as given. 1X just signed off on a 58,000 square foot floor in Hayward shipping ten thousand NEO units this year and a hundred thousand by 2027. Figure logged 1,250 hours at BMW Spartanburg moving ninety thousand parts across thirty thousand vehicles. Atlas units for 2026 already allocated to Hyundai and DeepMind. Xpeng's Iron rolling out of Guangzhou. GAC's GoMate behind it.

The first-year run sold out in five days.

What broke the frame was a constraint that decades of immigration politics, wage policy, and skills-gap consulting had been arguing along as though it were fixed. The labor curve they modeled assumed a worker function capital had no reason to keep paying for once the alternative cleared. None of the panels priced it. None of the unions priced it. Hayward is shipping ten thousand units anyway.

USA picks invention. China picks scale. Europe picks regulation. LATAM picks copying whatever Europe writes. Forty years of each population voting for the political class that delivers exactly that.

What Europe still exports at scale is regulation. GDPR was copy-pasted into the law of 75+ jurisdictions including most of LATAM. The AI Act is on the same trajectory. eIDAS is exporting biometric ID standards. Chat Control would scan every private message by default and gets reintroduced every legislative session. The European bureaucratic class regulates because that is what it gets paid for and what its voters reward. The median European voter prefers safety to growth, predictability to upside, the regulator to the entrepreneur. Nobody is forcing anyone.

Europe still produces. ASML controls EUV lithography for the entire global semi industry. BioNTech invented the mRNA platform Pfizer commercialized. Novo Nordisk reshaped the obesity drug market. Airbus and Boeing are a duopoly. Siemens and SAP still dominate their categories. The point is not that Europe stopped building. It is that the buildings are old. Most of these firms started before 1990, when the regulatory weight was a fraction of what it is today. The flow of new global champions has been near-zero for two decades. Mistral AI had to partner with Microsoft to serve global customers. Klarna chose NYSE for its listing. The stock pays dividends. The pipeline does not refill.

LATAM picked the worst available combination. Import the European regulatory template, skip the European invention base, add layers of local taxation on top. Capture every cost of the model and none of its flows. Argentina lived a complete cycle. Brazil is mid-cycle. Mexico looks next.

USA and China kept the lanes that produce, and they will keep producing. Europe and LATAM kept the lanes that consume what others produce, and they will keep falling behind in compounded terms while telling themselves it is a moral choice.

The voters chose. The voters can still unchoose. There is no sign yet that they will.

Bloomberg counts fifteen billion dollars spent on Starship like the number is a verdict. SLS burned more than that for one flight. Apollo burned more than that for hours of crewed time at the moon. The arithmetic that scores this as excess belongs to a procurement order trying to keep itself legible against an inferior unit cost.

Booster 19 ignited thirty-three Raptor 3 engines on April 15. Pad 2 took the load. Ship 39 sits stacked. Flight 12 opens on the May 12 window. Version 3 carries a hundred-ton payload target to low orbit, threading a corridor between Jamaica and Cuba that did not exist on any map produced by a national space agency.

The line that this is billionaire indulgence funded by taxpayers inverts the flow. NASA pays SpaceX a fixed-price milestone the legacy primes cannot meet at any price. Each launch shrinks the unit. Each failure folds back into iteration the state program would have folded into a Congressional hearing.

The capital has already decided what space costs. The rest is bureaucracies negotiating the speed of their own irrelevance.

Forty years of nuclear stagnation in the West was sold as caution. Lovins, Caldicott, the post-Three Mile Island licensing freeze, NRC's Linear-No-Threshold consensus rounding every microsievert toward apocalypse. The actual mechanism was capital mispricing the cost of fear while the grid drifted into a lattice of natural-gas peakers propped up by storage that did not exist at scale.

The Atomausstieg is the proof. Germany shut Isar 2 and Emsland with Russian tanks in Kharkov, bought Texan LNG at five times the marginal cost of the kilowatt it had just turned off, called the result a transition.

Now the bill matures. NRC is closing in on the first commercial SMR construction permits this year. Holtec is restarting Palisades, a plant the same regulator had buried. TVA is breaking ground at Clinch River on a BWRX-300. Rolls-Royce is exporting prefabs to Wales and Bohemia. Washington has filed for four hundred gigawatts by 2050, which would put the United States back where it was scheduled to be in 1985 before Carter blinked.

Demand is doing the work the lobbyists could not. Training campuses need three to five gigawatts of firm baseload that no wind farm in any latitude can deliver, and the hyperscalers can count. Microsoft signed for Three Mile Island Unit 1. Amazon bought Talen's Susquehanna. Google contracted Kairos. The capital allocators have stopped consulting the Sierra Club.

The soft path was a luxury good. It expired the moment a model on a TPU pod needed a real watt.

The lazy reading of Berkshire's $397 billion cash pile is that Buffett sees a crash coming. He doesn't, he never has, and he is bad at it. The real story is duller and more useful. T-bills at five percent are now the hurdle rate, and equity has not cleared it in fourteen consecutive quarters.

Cash up roughly 3x in four years. Apple cut by half in 2024. Bank of America trimmed from 13% of book to ~9%. None of that capital was redeployed in size. Fourteen straight quarters of net selling has no precedent in his career.

The arithmetic is plain. T-bills yield ~5% nominal, ~2.5% real. Any equity buy has to clear that plus a margin of safety, not match it. Buffett indicator at historical extremes. CAPE above 35. The position translates into "I do not understand the price," not "the market will fall." That distinction preserves optionality without committing to a forecast.

What makes this week different is convergence. Apollo's data on household ETF flows shows retail dumping long-duration Treasuries and piling into ultra-short. Berkshire is in the same trade for entirely different reasons. Smart money and dumb money rarely share a trench. When they do, the equity risk premium is usually too low and the room has finally noticed at once.

Capital costs went up, asset prices have not adjusted, the option to wait became valuable again. The anomaly was 2009 to 2022, not now. Cash earning real yield is the historical default. The market is just remembering.

The signal is patience. The richest balance sheet in the world is telling everyone else they do not have to play this hand.

The Senate's stablecoin yield compromise in the CLARITY markup is being sold as consumer protection. The function is older. Crypto issuers are blocked from passing T-bill yield through to holders, except where the transaction qualifies as "bona fide," a category whose job is to be ambiguous on demand. The carve-out is the bill. Regional banks fund mortgage duration with checking deposits paying a fraction of a percent while three-month paper prints over four; the spread is the franchise. A permissionless dollar that pays the spot risk-free rate directly to the bearer collapses that arbitrage on contact. Tether's billion-dollar quarter is the same number from the issuer side, kept rather than passed through. The statute freezes the spread by decree, on the implicit theory that you can legislate against a yield curve. The yield curve has not been notified.

DeepMind's new AI Policy paper quietly inverts a decade of framing about what science is missing. The hardware-talent-compute narrative was always the easier story. The actual bottleneck sits in institutional permafrost. Forty years of fusion data from JET, functionally inaccessible. Only ~10 fusion facilities worldwide. Validated training sets in the hundreds or low thousands of experimental shots. Releasing JET's archives requires consent from a stakeholder list nobody has bothered to assemble in twenty years. Simulation codes still run for weeks on supercomputers.

While the public science estate freezes at the data layer, $2.5B+ has flowed to 30+ private fusion companies in the last cycle. Capital moves. Data does not.

The paper names three debts. First, technical. Infra for data collection and curation has been under-funded relative to hardware for decades. Second, bureaucratic. Complex ownership webs and divergent open-data policies stall every release. Third, human. Too few software engineers and data scientists exist to clean, validate and curate. Postwar Big Science was built for an era when instruments and PhDs were the rate-limiting step. The AI era inverts the constraint. The new bottleneck is access to the exhaust of past instruments and the labor to make it legible.

This is Hayek's knowledge problem rewritten for physics. Information is decentralized, tacit, lives in facility-specific formats, sits behind consent regimes designed for analog publication. The system that produced the data cannot publish it without negotiating with itself.

The proposal is deliberately modest. Run "AI Data Stocktakes" per domain, expert audits that map gaps and convert them into fundable projects. Open-source 30% of JET data by 2028. Fund disruption-prediction competitions. Build a scientific data curation platform. Use AI agents to preserve expert knowledge encoded in legacy simulation codes.

Two readings collide here. The austrian read sees an inventory exercise to make existing capital stock legible to a new productive technology. The accelerationist read sees a slow public sector negotiating with a private compute frontier that already absorbed every available token. The friction between those two readings will eat most of the next decade's science policy.

The ITER reactor and the frontier training cluster are not the same machine. Each pretends the other does not exist. The stocktake is what happens when the pretense breaks.

https://www.aipolicyperspectives.com/p/science-needs-ai-data-stocktakes

The German social market economy had a deal. Robust state, dynamic productive base, the two compounding together. Erhard, ordoliberalism, the postwar miracle. The chart from Destatis is the obituary.

Public-sector headcount and self-employment crossed around 2008. Since then the blue line compounds at +1.7% a year. The orange line bleeds. Manufacturing shrinking five years running, chemical output at a 30-year low, factories at sub-72% utilization. Opening a business takes 120 days in Germany, twice the OECD average. Bundesagentur added 205,000 public hires in 2025 alone.

Diagnosing this as a management failure is the standard mistake. Mature welfare democracies generate this geometry by design. The median voter learns that public employment is the most direct way to vote himself a raise. The politician learns that hiring is faster than legislating. The would-be entrepreneur calculates the marginal cost of a German GmbH and stays an employee. Blue line has the incentive system, the ballot box and time on its side. Orange line has none of those.

Mises wrote Bureaucracy in 1944. The state does not produce. It redistributes. When the productive base stops growing fast enough to feed the apparatus, the difference is covered by debt, inflation or living-standard decay. Germany is now running all three. EUR 500B special fund, Bundesbank forecasting mediocre growth for years, a new chancellor promising "competitiveness" while signing off on more hires.

The chart looks technical. It is political. Each centimeter of divergence is another centimeter of jurisdiction gone from Erhard's republic to the administration's. The Germany of Volkswagen, BASF and Siemens cannot coexist with the Germany of Bundesagentur, AOK and Bürgeramt. The blue line feeds on the orange line. Destatis just put it in the public record.