This past week was a reminder that even the strongest trend in the market is not one-directional: IBM crashed after a pre-announcement, the chip index lost almost 9%, and above all one fear hovered - that the enormous investment cycle in AI infrastructure is starting to look excessive. This question will not be resolved by price action. It will be resolved by reading a chain of leading indicators, each of which measures a different part of the cycle. Here is the framework.
The Central Debate: Spending vs. Return
Before the signs, you have to frame the question correctly. The AI cycle is driven by two forces competing with each other: the pace of infrastructure investment (capex) versus the pace of return the investment generates (monetization). As long as revenue chases the investment at a reasonable pace, the cycle is healthy. The moment the investment accelerates far beyond the return - over time - is exactly the moment when historical cycles turned. All the signs below are really different ways to measure this ratio.
The Eight Signs
1. The hyperscalers' capex guidance. Google, Microsoft, Amazon and Meta are the demand side. Raising investment guidance = demand that continues; cutting or "digesting" capacity = cooling. This is the first number to look at in any report - and Alphabet, reporting this week, is a live test.
2. Supply-chain orders. What chip makers and equipment suppliers order today tells of demand a year from now. TSMC raised its 2026 capex budget to about $60 to $64 billion, and ASML also raised guidance - both are building capacity because customers demand it. A drop in orders is the classic leading sign of a slowdown.
3. The memory shortage (HBM). High-bandwidth memory is the tightest bottleneck in the AI chain. Shortage and price increases = strong demand (exactly what shifted IBM's customer budgets toward hardware and memory). An easing of the shortage would be one of the early signs of softening.
4. Power and land constraints. When companies report they are "running out of capacity" - as was reported about Google - it is a sign that demand exceeds supply. Energy and infrastructure constraints are today a real limiting factor, and a direct gauge of demand pressure.
5. Cloud revenue and margins. This is the most important sign of all, because it measures the return. If capex rises but cloud revenue and margins do not follow - that is a warning light for a bubble. Cloud growth together with margin expansion = the investment is translating into profit, and the spending is justified.
The gap that tells the whole story
Almost all the signs measure one side - the spending. The sign that separates a healthy cycle from a bubble is the other side: monetization. A company can raise capex forever as long as the market believes the return is on the way. The moment the gap between investment and revenue opens too much, and over time, the trust starts to crack - and that is the moment sentiment flips.
6. The chip makers' order backlog. The guidance and backlog of the data-center segments at Nvidia, AMD and others are the "picks and shovels" gauge of the race. A growing backlog = continued demand; a slowdown in the order pace usually precedes the slowdown on the ground.
7. Inference cost versus adoption pace. As the cost of running models falls and the pace of use rises, adoption expands beyond isolated labs into the broad market - a sign of the trend's maturity and health. If the cost stays high and usage struggles to expand, it hints that the value is not yet reaching the end user.
8. A red flag - circular financing. One of the most worrying signs in technology cycles is when demand is supported by circular financing: a supplier that invests in a customer, who buys back that same supplier's products. Such a flow can inflate reported demand that does not rest on independent economics. The larger the share of demand coming from such structures, the more caution should rise.
What This Week Tests
The current earnings season is a live laboratory for all these signs. Alphabet (Tuesday) will give a reading on both sides at once - the capex guidance versus the cloud margins. The reports of chip makers and chip users later in the week will complete the picture from the supply side. Whoever listens not only to the numbers but to the language of managements around the ratio between investment and return will get a far better indication than the chart tells.
The Bottom Line
The AI revolution is not a "yes or no" question. It is a cycle, and cycles have a pace. The only sign that really matters is the ratio between the pace of infrastructure investment and the pace of revenue it generates. As long as the two move together, the trend is alive. When the investment runs ahead and revenue stays behind over time - that is the sign the cycle is nearing a turning point. Whoever tracks this ratio, and not today's headline, sees the turn before it reaches the chart.
Sources: the data is based on ongoing coverage of earnings season (including Seeking Alpha, Zacks, Motley Fool, 24/7 Wall St) and on company reports, current as of the time of writing. Nothing herein is a forecast, recommendation or advice.