US chip export controls and the AI bubble
The essentials in 30 seconds
- As of October 2026, US chip export controls act on the AI bubble through two effects: lost markets and shifted supply.
- In a worked example, a restricted market worth 20% of sales cuts revenue by 20% and profit by 50%.
- The EQOS Asia AI study places the enterprise AI market at 0.337, a fragile zone (April 2026).
- Asia both buys AI hardware and hosts much of its supply chain.
- A four-step map shows where your chips, cloud, and models come from.
Chip export controls are rules that limit which advanced chips can be sold to which destinations. This page sits under AI regulation as a trigger for the AI bubble and covers the hardware side. It explains the two effects on an AI forecast, shows the arithmetic with round numbers, and gives a map of your own exposure. It is general information as of October 2026. Rules change often, so check the issuing authority for the current text.
On this page
Compare the EQOS Asia offers and request a 30-minute pre-diagnostic.
What are US chip export controls, and why do they matter to the AI bubble?
Export controls are government rules that decide which chips may be sold to which countries and buyers. Advanced AI chips are the base of the AI spending chain, so a rule on them reaches every level above. Prices of AI companies assume sales growth in many regions, and a rule can remove a region from that growth path.
A control is a policy decision and can change. The precise text is not summarised here, because it moves and the authority that issues it is the source to read. What lasts is the mechanism: a control changes who can buy, from whom, and at what cost. Those three changes are what an AI plan has to absorb.
How do chip export controls change AI revenue forecasts?
Forecasts add up sales by region. When a rule closes or narrows one region, its share drops out of the total. The company may sell a different product there, find new buyers elsewhere, or lose the sales. Each of those choices carries a different revenue path, and the forecast needs a plain assumption for each one.
Take an illustrative chip supplier with revenue of 100 and a 40% operating margin. A restricted market worth 20% of sales disappears. Revenue falls to 80. If costs stay at 60, profit falls from 40 to 20, a drop of 50%. A 20% revenue loss can halve profit when costs do not fall.
Which companies feel US chip export controls first?
Three groups feel a control first. Chip designers and makers lose part of their sales. Equipment suppliers lose orders from factories in the restricted market. Cloud and data centre firms that planned capacity around a specific chip must redesign their plans. Customers further up the chain feel it later, through prices and delays.
A second ring also matters: the companies that sell AI services built on that hardware. If hardware costs more or arrives later, projects cost more or slip. The EQOS Asia AI study records 95% of AI projects failing in production (EQOS Asia AI study, April 2026), so a further delay in hardware is a burden on a base that is already weak.
How can chip export controls shift AI supply across Asia?
Asia is both a large buyer of AI hardware and the home of much of its manufacturing and assembly. A control can therefore change orders between countries and companies in the same week. Buyers may move to other suppliers, other chips, or other regions, and each move changes lead times and prices.
For a company in Singapore, ASEAN, or Hong Kong the practical effects are indirect. Cloud prices, delivery dates, and model availability can all change. Treat the supplier of your compute as a risk to read each quarter, and note which chip generation your main providers rely on, since that tells you how a rule would reach your costs.
Do chip export controls burst the AI bubble or only reshape it?
In most cases they reshape. Demand moves from one buyer to another, and some companies gain what others lose. A control becomes a burst trigger when it removes a large part of expected sales from companies priced for strong growth, and no other buyers replace them. The test is replacement: who takes the orders?
Use the test on each company you follow. If replacement demand exists and is visible in orders, the effect is a reshuffle. If replacement is a hope, the forecast is at risk. The EQOS Asia AI study scores the market at 0.337, a fragile zone, so a hope-based forecast deserves a lower weight than one backed by orders.
How should you read a change in chip export controls over time?
A new rule passes through three stages. First comes the announcement, when prices react to the headline. Then comes the detail, when lawyers and companies learn what is covered and what is not. Last comes the effect on orders, which shows in company reports one or two quarters later. The first reaction is often larger or smaller than the final effect.
So separate the stages in your notes. Date the announcement, the published detail, and the first report that mentions an effect. Compare the three. A rule whose detail is narrower than its headline fades. A rule whose effect appears in reports across several companies in one season deserves a place on your watch list.
How can you map your chip and cloud exposure to export controls?
Use a four-step map. List your compute sources: own hardware, cloud, and model providers. Note the chips each one relies on and where they are made. Mark the sources that depend on a single region or a single chip. Add a back-up for each marked source, with its cost and the time to switch.
The result is a one-page table that shows where a rule would hit first. The Λ simulator from EQOS Asia lets you test a supply shock on your own case, and the EQOS Asia method turns the table into a score. A session with the EQOS Asia team applies it to a live decision. This page is general information, not investment advice.
See also in this guide
Frequently asked questions
What are US chip export controls?
They are government rules that decide which advanced chips may be sold to which countries and buyers. The rules change, so read the current text from the issuing authority. This page covers the mechanism, which stays stable.
Do export controls burst the AI bubble?
In most cases they reshape demand from one buyer to another. They become a burst trigger when they remove a large part of expected sales from companies priced for strong growth and no other buyers replace them.
How can a 20% revenue loss cut profit by 50%?
If costs stay fixed, the lost revenue falls straight into profit. In the worked example, revenue drops from 100 to 80 while costs stay at 60, so profit falls from 40 to 20.
What should an Asian company do first?
Map the sources of compute: own hardware, cloud, and model providers. Note which chips and regions each relies on, mark single points of risk, and plan a back-up with its cost and switching time.
Test your knowledge
1. What do export controls decide?
- Which chips may be sold to which buyers
- Share prices
- Office rents
Show the answer
Which chips may be sold to which buyers.
2. In the worked example, what happens to profit when 20% of sales disappear and costs stay at 60?
- It rises
- It falls from 40 to 20
- It stays at 40
Show the answer
It falls from 40 to 20.
3. What is the test that separates a reshuffle from a burst?
- The company logo
- The age of the CEO
- Whether replacement demand exists in orders
Show the answer
Whether replacement demand exists in orders.
4. Which groups feel a control first?
- Chip designers, equipment suppliers, and cloud planners
- Hotels
- Newspapers
Show the answer
Chip designers, equipment suppliers, and cloud planners.
5. What does the map of exposure mark?
- Favourite suppliers
- Single points of risk
- Office locations
Show the answer
Single points of risk.
About the author
Stéphane Jambu
President, EQOS Forecast Asia
Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile
Sources: EQOS Asia AI study (eqos-asia.com/case-studies/ia-2026.html), EQOS Asia method page (eqos-asia.com/method.html), consulted on 22 October 2026. Page updated on 22 October 2026.