Big Tech earnings must prove that extraordinary investment is creating durable economics rather than merely extending the spending cycle

The sell-off in semiconductor shares looks dramatic, but it does not yet establish that the artificial intelligence investment cycle is broken. The Philadelphia Semiconductor Index fell 10 per cent last week and sits more than 20 per cent below its June peak, yet remains over 60 per cent higher in 2026. This week's earnings should be read less as a referendum on whether AI is real and more as an audit of who is converting it into cash.

From scarcity to scrutiny

The first phase of the AI cycle rewarded scarcity. Compute capacity, advanced chips, memory and data-centre infrastructure attracted capital because demand exceeded supply. That resembles the steep growth stage of an industry life cycle, when volume expansion obscures differences in business quality. The market is now entering an early shake-out in expectations. Competitive intensity is rising, model costs are falling, and customers have more architectural choices. Revenue can continue growing while economic rents migrate elsewhere.

A semiconductor supplier may report excellent orders while its customer earns an inadequate return on the infrastructure those orders support. Hyperscalers may accept near-term margin pressure if new capacity strengthens cloud share, lowers unit costs and creates valuable recurring workloads. Key questions are whether utilisation rises, pricing holds, incremental gross profit follows, and cash returns ultimately exceed the cost of capital.

Alphabet is the test. It reports on 22 July, with investors focused on AI spending, cloud growth and the durability of search economics. First-quarter Cloud revenue grew 63 per cent, exceeded 20 billion dollars and carried backlog above 460 billion dollars. Alphabet also expects 2026 capital expenditure of 180 billion to 190 billion dollars. Search remains the revenue engine within Google Services, whose first-quarter operating income reached 40.6 billion dollars. The risk is that defending the mature cash engine requires progressively more capital while the new engine earns lower returns.

One label hides several cycles

Calling this a Big Tech rotation is analytically lazy. Alphabet, Tesla, Intel and the leading AI suppliers do not occupy the same competitive or life-cycle position. Tesla's automotive operation faces mature economics and intense price competition, while energy storage is growing and its robotics and AI ambitions remain venture-like. Tesla delivered 480,126 vehicles and deployed 13.5 gigawatt hours of storage in the second quarter, but earnings must show whether those volumes translate into better margins and cash conversion.

Intel presents a problem. Its foundry revival is closer to a capital-intensive turnaround than a conventional growth story. Intel Foundry reported 5.4 billion dollars of segment revenue, up 16 per cent, but that figure includes intersegment transactions and accompanied a 2.4 billion dollar operating loss. Investors need proof that manufacturing execution, customer commitments and utilisation can improve faster than depreciation and funding demands. The same AI boom can create scarcity economics for one chipmaker, reinvestment risk for another and margin pressure for the customer buying the chips.

The macro backdrop raises the hurdle. UKOil crude above 90 dollars and a ten-year Treasury yield near 4.55 per cent combine inflation risk with a less forgiving discount rate. Long-duration equities can tolerate high investment when terminal cash flows appear large and dependable. They become vulnerable when energy costs, financing costs and technological uncertainty rise together. This does not end the structural AI expansion, but it reduces the present value of distant profits and exposes weak links in the value chain.

Follow the cash, not the applause

The decisive disclosures will sit below revenue. Investors should watch depreciation, operating margins, capital commitments, free cash flow, backlog conversion and management's language on utilisation. Microsoft's March-quarter revenue rose 18 per cent and operating income 20 per cent, demonstrating that heavy AI investment can coexist with operating leverage. Yet its cost of revenue grew 22 per cent, faster than sales, reminding investors that genuine demand does not guarantee stable unit economics.

The market may be mispricing the rotation itself. The price action suggests money is moving within technology, from businesses valued on scarcity and distant optionality towards companies able to show monetisation, balance-sheet endurance and credible incremental returns. Strong earnings could revive semiconductor momentum, but a reflexive rebound would not settle the argument. The next phase will reward evidence rather than exposure.

The single variable to monitor over the next two years is the conversion of AI capital expenditure into incremental free cash flow. If cash generation begins to scale behind investment, the recent sell-off will look like a violent reset within a continuing growth cycle. If depreciation rises, utilisation disappoints and customers resist pricing, the industry will have reached a harsher shake-out. AI can transform the economy and still disappoint shareholders who paid too early for profits arriving too late.