How to Lose the AI Race — On Purpose
There’s no bigger geopolitical prize today than securing leadership in artificial intelligence. Countries that dominate AI will gain not only economic advantages but also national security superiority and global political leverage. Yet, against all logic, the latest round of tariffs all but guarantees that the U.S. becomes the most expensive place on earth to build the infrastructure needed to compete in this race.
The structure of these tariffs reveals the fundamental flaw. While policymakers claim they’ve shielded semiconductors, the reality is more complicated. Chips themselves aren’t the issue. AI datacenters rely on fully assembled systems — servers packed with GPUs, storage arrays, networking hardware — mostly imported from Asia, particularly Taiwan, South Korea, Malaysia, and Vietnam. Tariffing these imports is like sparing the wheat but taxing the bread, the sandwich, and the toaster.
By the time the U.S. builds out domestic capacity to produce these finished systems — a monumental task likely spanning much of the decade — global competitors will have surged ahead, deploying AI at scale and reaping the productivity gains that come with it. We will have effectively disqualified ourselves while congratulating each other on our industrial policy.
There was an alternative.
The administration gambled that other major economies would quietly accept U.S. tariffs and come to the negotiating table with concessions in hand. That was wishful thinking at best. Trade partners don’t respond to economic pressure by folding — they counterpunch. To his credit, Bessent saw this and tried to steer toward de-escalation. But China, recognizing a gift when it sees one, kept the pressure on, seeing the U.S. policy misstep as a strategic jackpot.
Advantage: Beijing.
With one policy stroke, we’ve managed to push up costs for critical AI infrastructure while helping China solidify its supply chains and draw regional allies closer into its economic orbit. But perhaps more alarming is the growing sense that this may not be an accident at all.
There’s something telling in Secretary Bessent’s recent public comments. He’s made it clear the administration is moving as fast as possible, knowing they’ll soon be bogged down by special interests. The mindset appears to be that the gravitational pull of bureaucracy is so powerful, they need explosive momentum to escape its orbit entirely. There’s no confidence that traditional channels — lobbyists, trade groups, chambers of commerce — would ever deliver the agenda they envision.
Maybe they’re right.
But it raises a critical question: could a more surgical approach have been just as effective, without the collateral damage of a full-scale trade war?
This philosophy also helps explain an even darker possibility: that the administration may be deliberately welcoming economic disruption to force rates lower and ease the debt burden. Nearly $8 trillion of U.S. government debt is rolling over this year. With yields stuck near cycle highs, refinancing this mountain of obligations is a growing problem.
So, what’s one way to bring yields down? Engineer a sharp enough slowdown to drive capital back into government bonds, pulling rates lower through sheer fear.
It’s a brutal strategy, but not without precedent. In The Forgotten Depression, Jim Grant recounts how U.S. policymakers in the early 1920s endured a deep, fast recession to purge inflation and reset the economy. Prices fell, wages contracted, output declined — but crucially, borrowing costs collapsed, easing financial strains and paving the way for recovery.
Of course, the economy of 1920 wasn’t globally intertwined or operating at digital speed. Trying to recreate that playbook today risks triggering second- and third-order effects that spiral far beyond policymakers’ control.
Yet, the parallels are difficult to ignore. Rather than crafting precise industrial strategies, the administration seems to have embraced disorder, betting that the ensuing chaos will tame inflation and suppress yields — even if it means hobbling U.S. competitiveness in the process. It’s a high-stakes gambling.
What’s the best case from here? Swift damage control. Fast-track trade agreements with allied Asian nations — Japan, South Korea, Taiwan, Malaysia, Vietnam. Aim for modest tariffs with true reciprocity, phasing to zero over time. That could stabilize supply chains and prevent further erosion of U.S. leadership.
To be clear, this isn’t an argument against tariffs per se. Done methodically, with gradual rollout and paired with deregulation, tariffs could have supported reshoring and nudged partners toward lowering their own trade barriers. And before this policy shift, U.S. tariff rates were already below those of many global peers — a position of strength we’ve now squandered.
Bessent, it appears, favored this more balanced approach. Unfortunately, it seems he lost the argument inside the administration. With any luck, market signals will force a rethink before the damage becomes irreversible.
The clock is ticking.
On a lighter note (if one can call it that), you must wonder how Bessent felt discovering the method behind the reciprocal tariff calculations. Reports suggest parts of the regime were drafted by a language model, leading to some creative outcomes — like tariffs on remote islands populated only by penguins. It’s a fitting metaphor: we’re using artificial intelligence to craft policies that undermine our ability to lead in artificial intelligence.
But beneath the absurdity lies a serious problem. Policies made by machines, without proper human oversight, mirrors the larger issue: we’re undermining our ability to lead in AI by making the economics of AI infrastructure prohibitively expensive.
The deeper risk is that this tariff escalation doesn’t just raise costs — it accelerates the global shift away from U.S.-centric supply chains. America’s technology leadership has long depended on its reliability as a trading partner and its openness to innovation. These new policies send the opposite message.
Markets are already picking up the signal. Hyperscalers and enterprise tech buyers face rising costs. Hardware inflation could creep back in. Cloud margins, already squeezed by AI buildouts, are under fresh pressure. And perhaps most critically, policy uncertainty is back — threatening to freeze capital formation at the moment we need it most.
In short, we’re fighting the wrong war with the wrong weapons. What’s needed isn’t scattershot protectionism, but strategic statecraft that understands AI leadership is about far more than where chips are manufactured. It’s about building the ecosystems, the talent pipelines, and the deployment capacity to win the race that matters most.
Our competitors understand this. So do our allies. The question is: Will Washington figure it out before it’s too late?
IMPORTANT LEGAL DISCLOSURES
CURRENT MARKET DATA IS AS OF 04/07/2025. OPINIONS AND PREDICTIONS ARE AS OF 04/07/2025 AND ARE SUBJECT TO CHANGE AT ANY TIME BASED ON MARKET AND OTHER CONDITIONS. NO PREDICTIONS OR FORECASTS CAN BE GUARANTEED. INFORMATION CONTAINED HEREIN HAS BEEN OBTAINED FROM SOURCES BELIEVED TO BE RELIABLE BUT IS NOT GUARANTEED.
THIS PRESENTATION (THE “PRESENTATION”) HAS BEEN PREPARED SOLELY FOR INFORMATION PURPOSES AND IS NOT INTENDED TO BE AN OFFER OR SOLICITATION AND IS BEING FURNISHED SOLELY FOR USE BY CLIENTS AND PROSPECTIVE CLIENTS IN CONSIDERING GFG CAPITAL, LLC (“GFG CAPITAL” OR THE “COMPANY”) AS THEIR INVESTMENT ADVISER. DO NOT USE THE FOREGOING AS THE SOLE BASIS OF INVESTMENT DECISIONS. ALL SOURCES DEEMED RELIABLE HOWEVER GFG CAPITAL ASSUMES NO RESPONSIBILITY FOR ANY INACCURACIES. THE OPINIONS CONTAINED HEREIN ARE NOT RECOMMENDATIONS.
THIS MATERIAL DOES NOT CONSTITUTE A RECOMMENDATION TO BUY OR SELL ANY SPECIFIC SECURITY, PAST PERFORMANCE IS NOT INDICATIVE OF FUTURE RESULTS. INVESTING INVOLVES RISK, INCLUDING THE POSSIBLE LOSS OF A PRINCIPAL INVESTMENT.
INDEX PERFORMANCE IS PRESENTED FOR ILLUSTRATIVE PURPOSES ONLY. DIRECT INVESTMENT CANNOT BE MADE INTO AN INDEX. INVESTMENT IN EQUITIES INVOLVES MORE RISK THAN OTHER SECURITIES AND MAY HAVE THE POTENTIAL FOR HIGHER RETURNS AND GREATER LOSSES. BONDS HAVE INTEREST RATE RISK AND CREDIT RISK. AS INTEREST RATES RISE, EXISTING BOND PRICES FALL AND CAN CAUSE THE VALUE OF AN INVESTMENT TO DECLINE. CHANGES IN INTEREST RATES GENERALLY HAVE A GREATER EFFECT ON BONDS WITH LONGER MATURITIES THAN ON THOSE WITH SHORTER MATURITIES. CREDIT RISK REFERES TO THE POSSIBLITY THAT THE ISSUER OF THE BOND WILL NOT BE ABLE TO MAKE PRINCIPAL AND/OR INTEREST PAYMENTS.
THE INFORMATION CONTAINED HEREIN HAS BEEN PREPARED TO ASSIST INTERESTED PARTIES IN MAKING THEIR OWN EVALUATION OF GFG CAPITAL AND DOES NOT PURPORT TO CONTAIN ALL OF THE INFORMATION THAT A PROSPECTIVE CLIENT MAY DESIRE. IN ALL CASES, INTERESTED PARTIES SHOULD CONDUCT THEIR OWN INVESTIGATION AND ANALYSIS OF GFG CAPITAL AND THE DATA SET FORTH IN THIS PRESENTATION. FOR A FULL DESCRIPTION OF GFG CAPITAL’S ADVISORY SERVICES AND FEES, PLEASE REFER TO OUR FORM ADV PART 2 DISCLOSURE BROCHURE AVAILABLE BY REQUEST OR AT THE FOLLOWING WEBSITE: HTTP://WWW.ADVISERINFO.SEC.GOV/.
ALL COMMUNICATIONS, INQUIRIES AND REQUESTS FOR INFORMATION RELATING TO THIS PRESENTATION SHOULD BE ADDRESSED TO GFG CAPITAL AT 305-810-6500.