Can AI burst its own bubble? What the global tech market could face in 2027

This is the first column by Samir Hajibayli, a VC of Caucasus Ventures. In his debut article, he examines three developments that could shape the tech market in 2027: growing concerns over AI safety, rising interest rates in the U.S., and changes in crypto market regulation.

Read the full article below.

Samir Hajibayli, Baku city, VC Caucasus Ventures, LinkedIn

AI safety warnings, higher US interest rates and a failed crypto vote are changing the outlook for 2027.

Three events shaped the week. AI researchers raised stronger concerns about safety. The Federal Reserve increased interest rates by 25 basis points. The U.S. Senate failed to advance a major crypto market structure bill. These events may seem separate, but they all raise the same question: what happens when trust becomes harder to earn? AI companies must show that they can control their systems and cover their rising costs. Central banks must keep inflation under control. Digital asset companies must prove that tokens represent real and legally protected assets. For the Caucasus and Central Eurasia, the main lesson is simple. We should not copy every new U.S. trend. We should understand which trends global capital will continue to support.

AI safety warnings may also help the largest companies

Anthropic researcher Jacob Coxon resigned on 8 September. He said leading AI laboratories were “gambling with our lives.” Joe Benton, a former leader of Anthropic’s safety team, separately warned that oversight was not keeping up with new AI capabilities. A few days later, Anthropic CEO Dario Amodei asked  leading laboratories to slow the development of their most powerful models. He did not call for all AI research to stop. He proposed permanent outside evaluators inside major laboratories, coordination between competitors and international rules before AI systems become more autonomous.

These warnings are based on events seen during testing. Some frontier models have moved beyond test limits, reached outside systems and acted in ways their developers did not expect. Amodei warned that groups of AI agents might be able to attack large parts of the internet within six to twelve months. No AI system has yet caused major public harm on purpose. Still, developers cannot fully explain or control every behavior. Software can spread quickly, so the lack of a disaster so far does not mean that current controls are strong enough.

The warnings may also serve the business interests of the largest AI companies. Strict rules require costly testing, security teams and licenses. Smaller competitors may not be able to afford them. Anthropic is also reportedly considering another powerful model before a possible public offering, even while Amodei asks the industry to slow down. This shows the basic problem. Each company may support limits for the whole industry but fear losing ground if it slows down alone. The safety concerns can be real even when new rules strengthen the largest companies.

Geoffrey Hinton’s recent statement has also been reported too strongly. He told U.S. lawmakers that they may have about a year to introduce better controls before advanced systems become much harder to manage. He did not prove that artificial general intelligence will arrive within one or two years. There is no common test for AGI, so its arrival date should be treated as a possibility rather than a forecast.

The AI bubble depends on costs and returns

Stanford’s 2026 AI Index estimates that global corporate investment in AI reached $581.69 billion in 2025. Private AI investment was $344.7 billion, including $170.9 billion for generative AI. At the start of 2026, 5 leading frontier AI companies had estimated annualized revenue of about $44.9 billion in total. On that snapshot, corporate investment was almost 13 times revenue, while private investment was almost 8 times revenue.

This comparison has limits. The investment figure covers a full year and includes financing and acquisitions. The revenue figure is an annualized sales rate measured at one point in time. Revenue is also growing quickly. By July, OpenAI and Anthropic reportedly had more than $105 billion in combined annualized revenue. Even so, revenue alone does not prove that the economics work. Companies still have to pay for computing power, customer growth and new models while prices are falling. The Financial Times reported that OpenAI expects almost $280 billion in cash burn from 2026 through 2030. AI can change the economy while still producing poor returns for investors who paid too much.

AI could weaken its own market in several ways. A serious safety incident could make AI slower and more expensive to deploy. Open and smaller models could push prices down faster than infrastructure costs fall. Enterprise customers may also stop paying if AI tools do not create lasting value. Safety rules could raise costs further, although those same rules may protect the largest firms from new competition. The most likely outcome is not the end of AI. It is a market correction in which useful products survive and costly experiments disappear.

The Fed has made capital more expensive

On 16 September, the Federal Open Market Committee voted 12 to 0 to raise its target rate to 3.75 to 4.00%. The hawkish message did not come from FOMC minutes, because they have not yet been published. It came from the official statement, the new Summary of Economic Projections and Chair Kevin Warsh’s press conference. The median participant now expects the policy rate to reach 4.1% at the end of 2026, compared with 3.8% in June. From the current midpoint, this suggests at least one more quarter point increase if the outlook does not change. 17-18 participants saw higher risks from inflation. Warsh put the message simply: “Inflation is the problem.”

Higher U.S. interest rates affect our region even when local central banks make different decisions. Higher U.S. bond yields can strengthen the dollar and increase the return that global investors expect. This makes emerging markets less attractive and pushes technology valuations down. It also creates a currency risk for companies that raise money in dollars but earn revenue in lari, dram, tenge, manat or som. Refinancing becomes harder, while imported technology and cloud services become more expensive in local currency.

The effect will not be the same in every country. Azerbaijan has energy income and a managed exchange rate, which provide some protection. However, tighter global liquidity can still affect reserves and bank lending. Kazakhstan also has commodity income, but its floating currency makes the adjustment more visible through the tenge and inflation. Armenia, Georgia and Uzbekistan have different links through remittances, trade and foreign funding. One fact applies across the region: the U.S. risk free rate sets the minimum return for many global investors. Founders should keep more cash, reduce currency risks and connect each new funding round to clear results. Investors should reserve more money for follow on rounds and avoid plans that depend on a quick return to cheap capital.

Crypto adoption is continuing under tighter rules

The Senate vote on 15 September was a failed procedural vote, not a rejection of blockchain. 50 senators voted to advance the CLARITY market structure framework and 49 voted against it. The bill needed 60 votes, so it did not move forward and is likely to remain stalled before the midterm elections. The debate covered more than transaction transparency. Senators also disagreed about conflicts involving public officials, stablecoin rewards, anti money laundering rules, the definition of decentralized finance, fundraising exemptions and the division of authority between the SEC and the Commodity Futures Trading Commission.

Four days later, SEC Chair Paul Atkins announced a five year Innovation Exemption for venues that trade tokenized U.S. listed shares. The exemption has clear limits. Access must be permissioned, sanctions rules still apply and synthetic stock tokens are not allowed. Each token must carry the same financial and voting rights as the underlying share. Laws against fraud and market manipulation also remain in force. This shows that tokenization can continue to grow, but mainly inside a regulated system. The exemption is less secure than a law because a future SEC or a court could narrow or reverse it.

The opportunity for the Caucasus and Central Eurasia is to build trusted systems for cross border payments, remittances, trade finance and tokenized real world assets. The region should not try to compete by offering weak supervision. Dollar stablecoins and U.S. rules will continue to shape global liquidity. Regional products therefore need identity checks, sanctions screening, secure custody, audited reserves, clear redemption terms and a legal claim to the underlying asset. A blockchain record alone does not create legal ownership.

What the region should do now

Entrepreneurs and investors in our region cannot ignore the United States. It still sets the price of dollar funding, hosts the leading AI companies and shapes the institutional digital asset market. But following global developments does not mean copying every trend. In AI, we should support capital efficient products with export revenue, useful private data and a clear paying customer. In finance, we should prefer companies that can survive a stronger dollar and a longer fundraising process. In blockchain, founders should build for several regulatory markets from the beginning and treat compliance as part of the product.

All three stories come back to trust. AI companies want society to trust systems that are becoming harder to predict. Central banks want markets to believe that they will control inflation. Tokenized assets require investors to believe that a digital token represents a real asset with legal protection. In each case, trust must be supported by controls, audits and enough capital to cover the risks.

The key question for 2027 is not the exact date of AGI, the timing of interest rate cuts or the return of one crypto bill. It is whether companies can create measurable value when funding costs more, rules become clearer and customers demand evidence. Before investing, we should ask one question: if cheap dollars, unclear regulation and optimistic forecasts disappear, does the business still work? The strongest companies in our region will be able to answer yes.