Who will win the AI race? Capital allocation may give clues

There are few irrefutable truths in investing, yet prudent capital allocation is almost always at the heart of success. Companies, industries, and countries that invest in high-return opportunities, while others hesitate, consistently outperform.

The Winners & Warnings series examines these dynamics in real time, identifying the conditions that shape tomorrow’s leaders. Guided by the capital allocation framework (CAF), we assess not only what companies do but also when and why. By tracking industry investment levels and returns on capital, we aim to capture the inflection points where leadership changes.

Historical parallel – the iPhone cycle

When the first iPhone launched in 2007, the race for suppliers was simple: Get designed in. A single design win with Apple meant access to one of the fastest-growing products in history, years of recurring revenue, and instant credibility.

During those first few iPhone generations, suppliers that secured key positions – whether in displays, processors, memory, or sensors – saw cash flow surge and faced three options:

1. Reinvest. Expand capacity, accelerate R&D, and get ahead of the next generation’s specifications.
2. Acquire. Broaden capabilities or diversify revenue.
3. Return capital. Return capital to shareholders instead of expanding.

The companies that reinvested created a self-reinforcing flywheel:

Early win → Higher cash flow → More reinvestment → More design wins → Even higher cash flow

In mobile cameras, leaders poured resources into better sensors and optics, keeping their iPhone slots while expanding into other flagships. Those that bought back stock or wandered into unrelated markets saw their edge fade and lost Apple contracts, ultimately losing their place in Apple’s supply chain.

The lesson: Winning early is only the start. Sustained advantage depends on how you deploy capital after you win.

The AI cycle – today’s first few generations

Artificial intelligence (AI) infrastructure is in its early generations, much like the iPhone market in 2007–2010. The race is to be designed in to the defining systems of the era: AI accelerators, high-performance data center racks, and the networks connecting them.

An early design win can secure years of recurring demand. But history shows the real gap emerges in what happens next:

1. Reinvest. Double down on performance, scale, and integration to stay ahead of rising technical requirements.
2. Acquire. Fill capability gaps to own more of the system’s value (“content”).
3. Return capital. Choose not to expand and risk falling behind in the next cycle.

Companies that reinvest in this stage will be the ones with the capacity, technology, and credibility to win in successive generations. As with the iPhone, capital allocation choices at this pivotal time will shape the leaderboard for years.

Where reinvestment is most urgent in AI

1. Memory. Model sizes and training datasets are surging, driving sharp increases in high-bandwidth memory per chip and server. Leaders are committing billions to next-gen specs, while underinvestors risk losing design wins.
2. Power. AI pushes performance-per-watt. Advanced power management demands continual innovation and manufacturing upgrades.
3. Fiber. High-speed optical connectivity, the backbone of AI networking, requires higher-spec fiber with each generation. Reinvestment secures both capacity and performance.
4. Advanced packaging. Chiplets, 3D stacking, and thermal constraints become the performance bottlenecks and demand capital-intensive scaling of capability and yield.

These are not optional upgrades; they are battlegrounds where capital allocation decisions now will define future market share and margins.

Today’s cycle – bigger, faster, more global

Unlike the iPhone era, AI cycles move at breakneck speed. A design win today may be obsolete in under a year, compressing the time it takes for capital allocation decisions to play out.

It’s also global. The supply chain spans continents, with each step potentially shifting to the player investing fastest and smartest. Leaders are already committing billions to secure manufacturing, upgrade capabilities, and dominate supply chains. Followers are struggling to match the previous generation’s benchmarks.

Conclusion

Capital allocation drives competitive advantage, and in AI, the stakes are higher than ever. Early winners face the same choice iPhone suppliers did in 2007: Reinvest and acquire, or return capital. History favors those who reinvest with focus and urgency.

The CAF is built to find these moments early, targeting the sectors where reinvestment pressure is highest and capital is flowing fastest, before the market fully prices in tomorrow’s winners.