History, sovereignty and the next leg for markets
“History is a gallery of pictures in which there are few originals and many copies.” – Alexis de Tocqueville
Prologue
A leadership transition is already underway, but much of the market remains anchored to the last trade. Investors are still crowded into semiconductors, memory and artificial intelligence (AI) infrastructure supported by passive flows and a sell-side that continues to extrapolate an era of outsized earnings surprises that has largely run its course. The earnings revision cycle in AI-adjacent infrastructure is maturing. The bottleneck trade is crowded, over-owned and increasingly exhausted at current valuations. Tocqueville’s line belongs here: what is being mistaken for novelty is, in fact, a familiar late-cycle pattern, consensus convergence, narrow leadership and risk concentrated in the assets investors perceived as safest.
This does not diminish the importance of AI—it changes how it should be expressed. Semiconductors now represent an unusually large share of the S&P 500, reflecting both genuine industrial significance and the depth of institutional positioning. A period of digestion is likely. The historic run of outsized earnings surprises is probably unlikely to repeat. For much of the semiconductor complex, the story from here is about disciplined execution against elevated expectations rather than another wave of multiple expansion. For memory, by contrast, there is still scope for genuine cyclical upside and renewed multiple expansion if pricing power and supply discipline persist.
The AI trade is not over—it is being repriced and de‑risked. The recent correction in high‑beta tech has removed excess primarily through price rather than earnings, leaving the broader index far from bubble territory. As the AI bottleneck trade, arguably the most crowded in market history, unwound, its correction was as breathtaking as the run‑up that preceded it. As in the 1990s, a powerful secular theme will be punctuated by repeated, often violent corrections, which are a feature rather than a bug of the cycle.
As AI matures, its expression in portfolios must become more focused and nuanced, shifting from blunt bottleneck plays towards genuinely differentiated exposures across hardware, software, data, and infrastructure. The Iran shock has injected volatility, but markets are treating it as acute rather than systemic, a risk‑off reset from stretched levels, not the start of a secular downturn. The underlying bull case, anchored in AI‑driven productivity, supply‑side investment, and a resilient U.S. economy, remains intact. This drawdown is better understood as an opportunity to rotate, upgrade quality, and trim parabolic exposures, not abandon the theme.
Meanwhile, something broader is happening beneath the surface. The median stock is delivering strong earnings growth, with second-quarter results tracking towards roughly 25 per cent year-on-year. This is not a narrow AI melt-up but a rolling recovery. Equal-weight indexes, small caps and domestic cyclicals are gaining traction, supported by improving fundamentals and still-muted positioning. In parallel, the policy architecture for the next leg of U.S. financial leadership is quietly being built, with initiatives such as the Digital Asset Market Clarity (CLARITY) Act signalling an intent to formalize U.S. leadership in crypto and digital market structure.
Investors can wait for the old trade to work again, or they can adapt to the new one. The market has already begun to choose.
With semiconductors now arguably the most crowded trade in market history, the risk‑reward has shifted. Further upside increasingly depends on flawless execution and sustained sentiment, while downside risk is amplified by positioning alone. Capital is beginning to look elsewhere, towards less crowded, earlier‑cycle, and more balanced exposures.
The real economy trade
As the U.S. marks its 250th birthday, it is worth remembering that the country’s deepest habit is not stability but rebirth. From the moment the Puritans walked away from Europe, they were making a wager that when institutions decay, one does not merely complain; one leaves, builds and begins again. The “city upon a hill” was never a victory lap. It was a challenge, a warning and an act of civilizational competition. If the experiment failed, the world would see it. If it succeeded, the world would see that too.
That pattern has shaped nearly every major American inflection point since. Manifest Destiny [1] turned the continent into a balance sheet of possibilities. Rails, mills, mines, ports and towns were not merely projects; they were proof that the republic could convert belief into productive capacity. Pearl Harbor followed the same script in harsher form. One surprise strike, one day of infamy, and the U.S. moved with astonishing speed from shock to declaration, then from declaration to industrial mobilization. America’s instinct, when hit, has historically not been paralysis—it has been construction.
That matters now because the market continues to underestimate how deeply that reflex is embedded in the American political economy. President Donald Trump’s politics sit squarely inside that tradition. “American carnage” is a bleak diagnosis, but it is not a concession speech. It is a prelude to “make America great again.” What is distinctive is the explicit reach back to Alexander Hamilton, Henry Clay and the American System: tariffs as strategic tools, internal improvements as national platforms and finance directed towards productive investment rather than abstract arbitrage. This is not simple nostalgia in patriotic dress. It is an effort to reload the operating system that powered the first American rise, this time in a world of chips, energy systems and tactical supply chains.
This is why the more important market development now is reallocation into the real U.S. economy. AI is important, but it is no longer the only game in town. Tax relief, investment incentives, deregulation and a more permissive attitude towards productive capital are beginning to generate visible earnings growth through channels far wider than data centres and accelerators. Industrials, financials, transport, energy and domestically oriented smaller companies are all beginning to respond.
It also changes how growth itself should be interpreted. The traditional Keynesian reflex is to treat strong nominal activity as an inflation problem waiting to happen. Supply-side economics starts from a different premise. Growth driven by more production and capital formation, and higher productivity is not the same as growth driven by consumption booms and fiscal sugar highs. The former expands capacity. The latter strains it. If policy raises the productive frontier, output can rise without embedding the same inflationary dynamic policymakers fear when demand alone surges.
That is the essence of the sovereignty argument now emerging in Washington. An America First approach to growth means reshoring industry, securing supply chains, encouraging domestic capital formation and reducing dependence on fragile global inputs. National strength is being redefined less as financial abstraction and more as industrial depth. Growth, in that frame, is not the thing to be restrained. It is the mechanism for debt sustainability, real wage gains and higher living standards.
And yet the Street still trades as if U.S. exceptionalism were a stale factor, fully priced, fully played and now in structural fade. That view does not fit the historical record. When America fuses grievance with a coherent program for rebuilding capacity, the result has historically not been gentle mean reversion. It has been inflection. If the Hamilton-Clay logic now acquires real policy teeth under Trump, investors are not watching the twilight of an exhausted hegemon. They are standing near the front end of another American up-cycle. That is why capital should be moving towards Real America.
AI after the bottleneck
None of this makes AI less consequential. It simply places it in its proper context. The enduring mistake is to describe AI as though it were merely the next software or app boom. In reality, it is becoming an industrial system: a metred flow of inference produced by chips, memory, electricity, cooling, networking and very large capital expenditure. But the easiest market expression of that idea—owning the bottlenecks in semiconductors, memory and AI power—has become crowded enough to distort both risk and valuation.
The first phase of the AI cycle revolved around training frontier models. That phase attracted headlines because it was spectacular, expensive and episodic. The more durable economics lie in inference: the continuous work of serving prompts, enterprise tasks and digital agents. Over a model’s life, inference can consume several times the compute, memory and electricity used in the initial training run. The long-run economics of AI are therefore governed less by launch events than by sustained operation. The important units are no longer just graphics processing units (GPUs) or model parameters, but tokens, watts, cost per token and revenue per token.
Now the story is shifting again, from models and infrastructure to AI agents, enterprise adoption and productivity leverage. Even at tens of millions of agent users, the market is still early. Agents are not merely another interface. They are increasingly embedded inside workflows, customer support, software development, logistics, research and enterprise operations in ways that amplify output rather than simply automate isolated tasks. The real prize is not novelty. It is leverage: more throughput from the same human and institutional base.
That is why the claim that there is “just not going to be enough compute in the world to satisfy all the demand” should be taken seriously. Inference at scale does not only require chips. It requires the entire system: power, memory, cooling, networking, capex and operational discipline across the full stack. As enterprise adoption broadens and agents become persistent rather than experimental, demand for compute may remain structurally stronger than the current equity fatigue in AI leaders suggests.
This industrial reality still creates major opportunities, but they are no longer cleanly expressed through the old scarcity trade. The era of massive upside earnings surprises in semiconductors and AI power is probably over. Future beats are less likely to shock the market because expectations already begin from elevated levels. The risk is now more perverse: the same names investors once treated as neutral toll roads are increasingly entangled in platform rivalry, pricing pressure and political scrutiny. Infrastructure is no longer just infrastructure. It is strategy.
There is also a constructive reason not to abandon the sector. Hyperscalers may begin generating AI revenue faster than the Street assumes. Enterprise adoption is moving from experiment to production. Usage-based pricing is becoming more visible. Health care is one promising arena: hospitals, insurers, pharmaceutical groups and clinical software providers sit on large pools of unstructured data and costly administrative workflows, making the sector fertile ground for AI-driven productivity gains in coding, imaging, triage, prior authorization and drug-development support. The AI cycle is not over. It is evolving away from the narrow bottleneck trade and into a more mature phase of monetization.
Three secular themes—one emotional cycle
At the broadest level, the current market is being driven by three powerful secular themes: AI, an America First industrial and sovereignty revival, and Bitcoin as a non-sovereign store of value. Each speaks to a different dimension of the same underlying story: technology as leverage, national capacity as strategy and monetary architecture as risk. None of these themes is likely to disappear in the next few quarters. They are multi-year projects, not passing fashions.
What changes, and must change, is the tactical weight assigned to each. Psychology still governs the path. Fear and greed are not embarrassing footnotes to an otherwise rational process; they are the mechanism through which capital adjusts to new information. When greed concentrates in AI bottlenecks and America First cyclicals are still treated as an afterthought, the discipline is to trim the former and build the latter. When fear compresses Bitcoin’s price despite clearer rules and a strengthening structural case, the discipline is to accumulate rather than join the capitulation.
In that sense, the portfolio is simply a reflection of emotional rebalancing. AI deserves a structural allocation because it will shape productivity and corporate earnings for years, but it no longer deserves the same crowded, linear bet it did when the story was young and the weights were light. America First deserves a structural allocation because the policy regime is tilting towards domestic capacity and supply-side growth, but it will move through cycles of euphoria and disgust that must be traded around. Bitcoin deserves a structural allocation because it remains the cleanest non-sovereign store of value, but it will keep oscillating between being treated as salvation and being treated as a joke.
The task for investors is not to rise above emotion, but to recognize where the crowd’s emotion has gone and do the opposite at the margin. AI, America First and Bitcoin are all strong secular themes. The question is not whether they matter. It is how much they should weigh at any given moment when fear and greed have already done some of the work.
What should investors do?
The investment conclusion is not to choose between AI and the real economy, but to recognize that the market is reallocating towards both in a more selective way. The old approach—crowd into semiconductors, memory and AI power, and assume scarcity will do the rest—is no longer sufficient. AI remains a strategic technology and an important source of future productivity growth. But it now needs a rest from being treated as the only credible source of earnings momentum in the entire market. The evidence increasingly suggests otherwise.
For capital allocators, that means distinguishing between temporary scarcity rents and durable control points. Taiwan Semiconductor Manufacturing Company Limited (TSMC) remains critical because its wafer decisions govern leading-edge supply across the compute stack. Memory still matters because inference intensifies dynamic random-access memory (DRAM) and high-bandwidth memory (HBM) demand. Hyperscalers matter because they may monetize infrastructure faster than expected. But so do industrials, financials, transport, energy and domestically oriented firms that benefit from a hotter, more supply-driven U.S. economy.
The macro backdrop may soon reinforce that rotation. Kevin Warsh, Chairman of the U.S. Federal Reserve, does not sound especially alarmed about inflation in the old reflexive sense. Inflation has softened more decisively than many expected, policy still looks restrictive in real terms, and the Federal Reserve appears closer to easing than tightening, even if it refuses to advertise the fact. If that reading is right, a cross-asset momentum regime shift becomes plausible. Capital concentrated in narrow AI bottlenecks may begin rotating towards equal-weight equities, small caps, domestic cyclicals and alternative stores of value that benefit from easier financial conditions and a broader earnings base.
This is also where digital assets return to the picture, and where investors should stop treating policy as background noise. The Clarity Act should now be understood as passed, meaning the U.S. is no longer debating whether digital assets belong inside the financial system. It is writing the rules for how they will be integrated into it. Regulatory settlement changes the asset class from a perpetual political question into a field of capital formation, market structure and strategic competition. Digital assets are moving from the edges of finance into its operating system.
U.S. Treasury Secretary Scott Bessent captured the larger point at the New York Economic Club when he argued that standards are strategy. The country that sets the standards for digital assets, stablecoins, tokenization and new payment rails will enjoy disproportionate influence over the architecture of the next financial system. Trump’s stated goal that America should be the world leader in digital assets deserves to be taken literally, not dismissed as campaign ornament. In strategic terms, it sits alongside semiconductors, AI infrastructure and energy security as part of the same sovereignty agenda. The contest is no longer just about factories and fabs. It is about protocols, ledgers and the rules of digital value.
Within that framework, the base layer must be Bitcoin. Stablecoins matter because they extend dollar functionality and embed U.S. money more deeply into global transactions. Tokenization matters because it modernizes financial plumbing and broadens access to capital markets. But Bitcoin is the foundational digital reserve asset: scarce, neutral, global and outside discretionary monetary supply. If the U.S. is serious about leading in digital assets, it cannot treat Bitcoin as incidental to that ambition. It is the reserve substrate beneath the broader stack.
That is why the asymmetry is becoming more interesting. AI is priced as inevitability; Bitcoin is still priced by many as a speculative afterthought. Yet policy is moving in the opposite direction. Regulatory clarity, the emergence of a strategic reserve and the administration’s insistence on digital-asset leadership all push towards legitimacy, integration and state-level recognition. In that setting, dismissing Bitcoin is no longer caution. It is a failure to recognize where standards are being set and how those standards harden into strategy.
In this market’s latest phase, the AI-bottleneck trade has sucked the oxygen out of the Bitcoin trade. Capital, attention and narrative bandwidth migrated to the most crowded AI winners. Bitcoin’s thesis did not break; it was simply starved of air while the market chased what had already worked. Now, as the bottleneck trade saturates and policy settles around digital assets, that oxygen is starting to flow back.
The market texture is beginning to reflect that shift. Shorts are getting greedy. Volatility is compressed. Bad news is no longer breaking price in the same way. Those are often the first signs of a bottoming process: crowded bearish positioning, narrative exhaustion and a market that no longer decreases on information that previously would have pushed it sharply lower. Bottoms are rarely obvious in real time, but they often begin with precisely this change in behaviour.
The contrast with AI is striking. AI is priced as inevitability. Bitcoin is priced as irrelevance. One must continue to exceed already elevated expectations simply to justify its multiple. The other need only persist. That is what makes the asymmetry compelling. Consensus has loaded itself into the trade where upside must keep outrunning perfection, and neglected the asset where survival alone may be enough to force a repricing.
This is the allocation implication of the larger historical argument. If America is entering another phase in which crisis is converted into capacity, grievance into policy and policy into productive rebuilding, then capital should be positioned accordingly. That means owning the strategic layers of AI without worshipping the most crowded bottlenecks, owning the sectors tied to domestic industrial revival and remaining open to under-owned stores of value that stand to benefit when crowded narratives begin to crack.
This moment should be read as a rebalancing, not a repudiation. AI still matters enormously. But it now sits inside a wider market and economy that are responding to supply-side growth, deregulation and capital reallocation. Tocqueville’s insight still holds at the end as it does at the beginning: “History is a gallery of pictures with few originals and many copies.”
Today’s market has once again mistaken crowded consensus for safety and neglect for irrelevance. Winning investors will do the opposite.
[1] – Belief in the 19th century that the U.S. was destined to expand its culture and institutions across North America.
