Wu Fang-Ming's Perspective: The AI High-Stakes Bet Enters Act Two – Market Focuses on 'Burning Through Cash Flow'

In the second quarter of this year, two of the world’s most representative technology companies—Alphabet (the parent company of Google) and Tesla—simultaneously released their financial reports. Their revenue performance was nothing short of dazzling, yet both stocks fell sharply, highlighting how AI-driven tech stocks are now under intense market scrutiny.

Record-breaking revenue no longer guarantees upward valuation; instead, the AI race has evolved from a growth story into a financial endurance contest.

Alphabet achieved quarterly revenue of $119.8 billion, a 24% year-over-year increase, marking 12 consecutive quarters of double-digit growth. Google Cloud revenue surged 82%, with AI-related businesses continuing rapid expansion. Tesla also delivered strong results: $28.2 billion in revenue, 480,000 vehicle deliveries, and a 26% revenue growth rate—the best pace in nearly three years.

Yet, capital markets did not applaud. Following the earnings release, Alphabet’s stock plunged nearly 5% in after-hours trading, while Tesla dropped over 4%. Behind this contradiction lies a deeper market perspective: beyond 'selling the growth story,' investors are collectively expressing the same sentiment toward AI giants—not fear of stagnation, but growing concern over whether 'cash' will hold up.

The market is no longer focused on revenue, but on free cash flow.

A closer look at the financials reveals that Alphabet recorded negative free cash flow for the first time since its IPO in a single quarter. Tesla, too, turned negative for the first time in two years. Both companies reveal a critical truth: even the world’s most profitable tech firms are now seeing their cash reserves consumed by AI.

AI is transforming from a technological revolution into a capital endurance race. The market isn’t afraid of investment—it’s beginning to question: will this money ever be recouped?

For years, Wall Street has forgiven tech companies for burning cash. Amazon operated for years without profit; Netflix expanded for decades on negative cash flow; and Tesla completed the electric vehicle revolution while still in the red. In the past, as long as the vision was strong enough, markets were willing to wait. But today, the rules of the game are changing.

After all, the capital required for AI dwarfs any previous tech revolution.

In Q2, Alphabet’s capital expenditures reached $44.9 billion—nearly double year-over-year. Its full-year capex guidance has already been revised upward twice, now ranging between $195 billion and $205 billion. Tesla’s annual AI-related investment exceeds $25 billion, with the vast majority funneled into immature ventures like the Dojo supercomputer, robotaxis, autonomous driving, and the humanoid robot Optimus.

The issue isn’t investment itself, but that spending is expanding far faster than cash generation. As a result, Alphabet’s free cash flow turned negative at -$5.9 billion in Q2; Tesla’s dropped from $1.4 billion in Q1 to -$1.1 billion in just one quarter.

More alarmingly, Alphabet had signed future procurement commitments totaling $811 billion by the end of June—up nearly $500 billion in just three months—covering AI infrastructure needs such as chips, data centers, power, and networking equipment. This doesn’t even include the company’s annual capex budget of nearly $200 billion. In other words, Google isn’t just spending money—it’s locking in resources for the next several years.

The AI race has shifted from who has better technology to who can secure computing power, chips, energy, and data centers first.

The biggest cost of AI isn’t GPUs—it’s uncertainty

What’s truly making markets anxious isn’t capital expenditure, but the time to return on investment.

Google’s Gemini processes tens of billions of tokens daily, and cloud orders have surpassed $500 billion—impressive on the surface. Yet the company still had to admit, 'returns are still in the very early stages.'

Tesla’s case is even clearer. Robotaxis are only in trial operation in a few cities; the Optimus production line is still undergoing repeated adjustments; and while FSD (Full Self-Driving) paid subscriptions continue to grow, the annual revenue they generate is a mere fraction compared to the $25 billion invested in AI.

Markets are beginning to realize that AI is fundamentally different from any past technology product. Previous products could be sold upon completion, but AI infrastructure cannot. From data centers and GPUs (graphics processing units) to supercomputing, model training, and the formation of a stable business model, it could take three to five years—or even longer.

In other words, today’s massive investments are merely buying a ticket to the future. Whether that ticket ends up being first-class, no one knows.

Wall Street is shifting from 'fear of missing out' to 'fear of holding on'

The AI market rally over the past two years and into the first half of this year was essentially a FOMO (Fear of Missing Out) phenomenon. Markets feared missing the AI revolution and were willing to assign high valuations to any AI narrative. But market sentiment is now undergoing a fundamental shift. Investors are increasingly concerned about whether massive capital expenditures can be converted into free cash flow.

Goldman Sachs reports that global AI-related corporate bond issuance has reached nearly $490 billion this year, with 40% coming from hyperscale cloud companies. At the same time, credit spreads for hyperscalers are widening rapidly, indicating that markets are demanding higher risk premiums.

The reason is simple: AI requires not just equity capital, but also debt capital. But in a high-interest-rate environment, borrowing is becoming increasingly expensive, forcing capital markets to recalculate. If the future return on AI investments falls below today’s cost of capital, then the more invested, the greater the financial pressure created.

This is why, after the earnings release, the market no longer focused on Google Cloud’s 82% growth or Tesla’s record deliveries, but on both companies’ free cash flow turning negative simultaneously. It signals that the market believes AI’s bottleneck may not be technological—but financial. As a result, investors are selling stocks, shifting from 'fear of missing out' to 'fear of holding on' (FOHO).

The AI race is becoming a new 'arms race'

Alarmingly, Google and Tesla are not isolated cases. Meta, Microsoft, and Amazon continue to raise AI capital expenditures. Reports suggest ByteDance is considering raising its 2026 capex to $70 billion and seeking around $20 billion in new financing. In short, global tech companies are investing in AI infrastructure at an unprecedented pace. No company dares to slow down. Everyone fears that if they do, competitors will gain a decisive lead.

This psychology closely resembles the Cold War arms race. It’s not because more weapons are needed today, but because no one dares let others have more. As a result, every company is compelled to spend more—even as short-term returns vanish, cash flow deteriorates, and balance sheets face growing strain.

This is the most dangerous aspect of the AI arms race: not that companies want to burn cash, but that none have the option to stop.

DeepSeek’s 'philosophy of restraint' offers an alternative

Amid the global tech giants’ frenzied expansion, DeepSeek founder Liang Wenfeng’s remarks stand in stark contrast. His core message is simple: 'Restraint is a strategy.' In his view, the AGI market is large enough—there’s no need to rush to capture every segment. What matters is concentrating limited resources on the most promising direction. As such, DeepSeek has deliberately abandoned high-cost areas like video generation and super apps, focusing all its computing power, talent, and capital on foundational model research.

This is a fundamentally different strategy from Google, Meta, or Tesla. Not all-out war, but focused breakthroughs. Not chasing scale, but maximizing success probability. This doesn’t mean large companies are wrong.

Google must build data centers because billions of users worldwide need computing power daily. Tesla must develop robotaxis and Dojo because Musk is betting on a decade-long industrial transformation.

DeepSeek reminds the market of another truth: the AI race may not just be a capital contest, but an efficiency race. The biggest spender won’t necessarily be the ultimate winner. What truly determines victory isn’t how much money is burned, but who can establish a positive cash flow cycle the fastest.

Historical tech revolutions have never been about who spent the most, but who first built a sustainable business model. The internet, smartphones, and cloud services followed this pattern. AI will be no exception.

Today, Google, Tesla, Meta, and Microsoft are investing hundreds of billions of dollars—essentially buying the future. But markets are already asking sharper questions: First, if AI is truly so disruptive, why is free cash flow worsening? Second, if AI truly brings massive business opportunities, why must companies keep borrowing? Third, if AI truly creates an efficiency revolution, why is financial efficiency declining? These questions,