Big Tech just did something it had never done. Alphabet reported its most profitable quarter ever. Yet cash flowed out faster than it came in. Shares tumbled nearly 7%. The message from Wall Street came through loud and clear. Investors don’t like what they see in the artificial intelligence race.
This shift marks a departure from years of disciplined spending. Companies like Microsoft, Meta, Amazon and Google parent Alphabet once kept capital expenditures well below the cash their operations generated. No longer. The push to build massive data centers packed with graphics processors and specialized chips has flipped the script. And the numbers keep climbing.
Fortune detailed the milestone in a report on July 23. Alphabet hit $112 billion in profit, much of it from paper gains on stakes in SpaceX and Anthropic. Core operations showed strength too. Cloud computing revenue jumped 82%. Still, the company turned cash flow negative for the first time. Purchase commitments topped $800 billion. That includes $51 billion in backstops for other firms’ data centers. Management warned that 2027 capital spending would rise significantly higher.
Investors responded with sharp cuts to price targets. Piper Sandler, UBS and D.A. Davidson all lowered theirs. D.A. Davidson set a $350 target, one of the more bearish calls. Barclays stood alone in raising its target. The selloff rippled. Microsoft, Amazon, Nvidia and even Tesla felt the pressure. Tesla’s own high spending and margin squeeze sent its shares down 15%.
Gil Luria leads technology research at D.A. Davidson. He called the negative cash flow a “negative milestone.” People reacted, he said, because they sensed the industry would never reach this point. But Luria believes the market overreacted. The concern centers on valuation more than spending itself. At around $320 a share, he sees Alphabet valued as one winner among several. Not the only one.
Longer term, Luria expects profits to flow from supplying AI compute power. Alphabet could earn $15 billion to $20 billion this year from that activity. Google Cloud growth looks “out of this world,” he added. Nobody wants to hear optimism on a day when the chief financial officer signals things will get worse before they improve. Still, the fundamentals point to sustained demand.
Recent analysis backs up the scale. Reuters examined consensus estimates on July 22. U.S. hyperscalers could spend more on capital expenditures than they generate in free cash flow by 2027. Oracle already saw capex reach 174% of operating cash flow in its latest fiscal year. The group includes Microsoft, Alphabet, Amazon, Meta and Oracle. Their combined outlays now pressure cash generation as energy costs and hardware prices climb.
Goldman Sachs offered a broader forecast late last year. The bank projected AI companies might invest more than $500 billion in 2026. Consensus estimates for hyperscaler capital spending reached $527 billion, up from $465 billion earlier in the earnings season. Those figures have only grown since. Some projections now put 2026 spending near $700 billion or higher. By 2027, analysts at Evercore and Bank of America see the total topping $1 trillion.
Amazon’s guidance stunned many. The company signaled plans for $200 billion in capital expenditures this year. That dwarfs prior expectations and exceeds the entire U.S. energy sector’s annual investment. Alphabet followed with a $175 billion to $185 billion range. Microsoft eyes more than $120 billion in its fiscal 2026. Meta committed to significantly higher than its prior $70 billion level, with some estimates landing between $115 billion and $135 billion. Oracle targets $50 billion, a 136% jump.
These sums reflect real constraints. Microsoft disclosed an $80 billion backlog of Azure orders it cannot fulfill due to power shortages. Demand outruns even aggressive construction. Component prices have inflated budgets too. Microsoft attributed about $25 billion of its increase to higher memory, GPU and CPU costs. Meta pointed to similar pressures adding $10 billion.
Wall Street’s unease runs deeper than quarterly numbers. David Russell serves as global head of market strategy at TradeStation. “Earnings growth may not be enough to justify investment if capex is depleting cash,” he told Reuters. “Companies exist to make money, not spend money.”
Yet the spending fuels tangible growth. Consumers and businesses now direct $120 billion annually toward AI services. That figure stood near zero two years ago. Luria highlighted this as genuine demand. Not circular funding within a closed group of tech giants and startups. The ecosystem includes tight partnerships. Google, Amazon and Anthropic. Microsoft and OpenAI. Nvidia and CoreWeave. Some worry these arrangements resemble off-balance-sheet vehicles that masked risks in past booms. Luria flagged potential Enron-like exposures in those backstops.
Comparisons to historical technology cycles offer perspective. Goldman Sachs noted AI capex recently equaled 0.8% of U.S. gross domestic product. Past booms, including the late 1990s telecom surge, saw peaks above 1.5%. Reaching $700 billion in 2026 would align with those highs. Returns on invested capital appear to rise alongside the outlays so far. That suggests value creation rather than pure waste.
Still, free cash flow compression raises questions. JPMorgan Asset Management tracked the shift. AI-related capital expenditures claimed 33% of hyperscalers’ operating cash flow in 2023. Estimates put that share at 93% for 2026. Cash piles have dwindled. Some firms explore alternatives. Oracle signed $29 billion in capital-light deals where customers supply their own hardware or prepay. The company plans to fund roughly half its 2026 needs through equity issuance. Meta reportedly cut 10% of its workforce and eliminated 6,000 open roles to help offset costs.
Credit ratings agencies largely shrug off immediate danger. S&P Global Ratings reviewed the hyperscalers in May. Most balance sheets retain headroom. Alphabet and Microsoft benefit from diverse revenue and strong cash flow. Amazon and Meta operate with tighter but adequate flexibility. Oracle stands apart with higher leverage and weaker internal cash generation. Its credit profile looks more vulnerable to debt-funded expansion.
The infrastructure buildout already delivers winners beyond the hyperscalers. Nvidia reported data center revenue of $93.7 billion for its latest fiscal year, up sharply. Suppliers of memory, networking gear, power systems and cooling technology ride the wave too. Yet volatility has increased. When capex guidance rises without matching revenue upgrades, stocks suffer. The reverse holds when both move higher.
Some see Meta charting a different path. Reports suggest the social media giant may launch a cloud business to sell excess AI compute to outsiders. That could convert internal spending into new revenue. Its shares rose more than 9% on one such rumor in early July. At 18 times forward earnings, Meta trades as the cheapest among major technology names.
Power remains the binding constraint. Data center projects now compete for gigawatts in regions with limited grid capacity. Utilities scramble to add generation. Some hyperscalers explore nuclear restarts, small modular reactors and long-term renewable contracts. These efforts add to capital needs in the near term.
Analysts debate sustainability. Morgan Stanley and others have repeatedly lifted 2026 and 2027 forecasts. What began as $250 billion in expected AI-related spending for 2025 ballooned past $400 billion. Upward revisions continue. Goldman Sachs warned in a report titled “More AI capex, more volatility” that current 2027 estimates may prove too low. Growth could stay elevated rather than taper sharply.
History offers mixed lessons. Technology booms from railroads to fiber optics delivered oversupply and painful corrections. They also laid foundations for decades of productivity gains. The AI wave differs in speed and scale. Training a single frontier model now consumes energy equivalent to thousands of households for months. Inference demands multiply as agents and multimodal applications proliferate.
Executives insist the investments position them to capture massive markets. Sundar Pichai, Alphabet’s chief executive, has emphasized the transformative potential. Similar rhetoric comes from Satya Nadella at Microsoft and Mark Zuckerberg at Meta. Their actions match the words. Capacity shortages persist despite hundreds of billions deployed.
Short-term pain seems likely. Higher depreciation, energy bills and interest expenses will weigh on reported profits. Free cash flow may stay depressed into 2028 or beyond. Debt levels could rise for some players. Equity issuance dilutes ownership. Yet the alternative, ceding leadership in foundational models or cloud infrastructure, carries greater perceived risk.
Recent trading reflects this tension. Shares of hyperscalers and their suppliers swing on every capex comment. Nvidia, once a pure beneficiary, now faces questions about whether demand justifies its valuation if end customers struggle to monetize. Chipmakers and equipment providers post strong results but warn of potential pauses if Big Tech slows.
Luria maintains perspective. The $120 billion in real AI spending by end users validates the buildout. Businesses adopt generative tools for coding, customer service, analysis and content creation. Early returns appear promising even if broad productivity data lags. Consumer applications from chatbots to image generators drive usage.
The coming earnings season will test narratives again. Alphabet reports first among the group. Investors will probe for signs that cloud growth can accelerate further. They want details on monetization timelines for AI features. Similar scrutiny awaits Microsoft on Azure consumption, Amazon on AWS margins and Meta on advertising efficiency amid its AI experiments.
One truth stands out. The industry has crossed into uncharted financial territory. Never before have these cash-rich giants consistently spent beyond their means to chase a single technology. Wall Street’s reaction shows discomfort with the uncertainty. Whether that discomfort proves prescient or premature will shape returns for years ahead.
So the bets mount. Data centers rise across the country and globe. Power contracts multiply. Talent wars intensify. And the cash continues to flow outward. The question isn’t whether Big Tech can afford this sprint. It’s whether the returns will justify the strain once the infrastructure stands ready.
WebProNews is an iEntry Publication