Zoho founder and chief scientist Sridhar Vembu has issued a fresh warning about the massive investment wave surrounding artificial intelligence, saying the current AI credit boom could eventually meet a fate similar to the telecom investment bubble of the late 1990s and early 2000s. At the same time, Vembu has made an important distinction: he does not believe AI itself is a failure. Instead, he is questioning whether the scale of investment and credit flowing into the sector can continue at its current pace.
Vembu’s comments come at a time when technology companies, investors and governments are committing enormous resources to AI infrastructure. The spending is not limited to AI models and software. It is also creating demand for advanced processors, memory, data centres, electricity, cooling systems, transformers and other infrastructure required to operate large-scale AI systems.
According to Vembu, the current investment cycle has been supported by a substantial expansion of credit following the pandemic. In his view, this has created distortions across parts of the technology market, with AI-related demand affecting industries that may appear unrelated to artificial intelligence at first glance.
One of the most visible consequences, he suggested, is the rising cost of consumer technology. Smartphones and laptops have become more expensive, and Vembu connected part of this pressure to the enormous amount of investment being directed towards AI infrastructure. The competition for components such as memory and semiconductors is affecting both AI companies and manufacturers of everyday electronic devices.
Vembu’s argument is not that AI has no future. Instead, he believes the technology can become widely adopted even if the financial boom surrounding it eventually experiences a major correction. This distinction is central to his warning because technological progress and investment bubbles do not always move together.
The Zoho executive compared the present situation with the telecom boom of the late 1990s. During that period, companies and investors poured huge amounts of money into telecommunications infrastructure, particularly fibre-optic networks, as expectations for internet connectivity grew rapidly.
The technology itself eventually proved extremely valuable. High-speed connectivity became a fundamental part of the modern economy. However, the investment boom did not benefit every company that participated in it. When the telecom bubble burst in the early 2000s, numerous businesses suffered major losses even though the underlying technology continued to expand.
Vembu believes a similar distinction could apply to artificial intelligence. AI could become an essential technology across industries while some companies and investments associated with the current boom could still fail to deliver the returns expected by investors.
In other words, the success of AI as a technology does not automatically guarantee that every AI-related investment will succeed. Companies can build valuable technologies while investors simultaneously overestimate how quickly those technologies will generate profits.
This is an important issue because the current AI ecosystem involves enormous capital requirements. Large AI models require specialised computing infrastructure, powerful processors and significant electricity consumption. Building and operating data centres also requires cooling systems, power equipment and other supporting infrastructure.
As demand rises, the pressure can extend throughout the supply chain. Vembu has pointed to memory, CPUs, GPUs, electricity generation, transformers, backup generators and cooling equipment as examples of areas affected by the expansion of AI infrastructure.
Memory has become one of his particular concerns. Vembu has claimed that memory prices have risen sharply over the past year, describing an increase of around 500 per cent compared with the previous year and saying prices are several times higher than their lowest levels. These figures reflect his own assessment rather than an independently established industry-wide benchmark, but they underline his broader argument about supply pressure created by AI demand.
For software companies, rising infrastructure costs can create a difficult situation. Businesses that previously relied on relatively inexpensive computing resources may now have to account for significantly higher expenses related to memory, computing and AI services.
Vembu has said that Zoho has so far avoided passing all of these increased costs on to customers. However, he has also indicated that absorbing rising expenses indefinitely could become difficult.
The changing economics of memory also raise questions about software development. For decades, programmers have generally benefited from increasing computing power and declining costs of memory and storage. Software could therefore become more resource-intensive without necessarily creating a major increase in the cost of running it.
Vembu believes that assumption may now need to be reconsidered. If memory becomes significantly more expensive, developers may have to pay greater attention to how efficiently their software uses available resources.
This could encourage renewed interest in memory-efficient programming languages, better compilers and software that achieves more with fewer computing resources. The development of AI itself could also be affected because advanced AI applications can require substantial amounts of memory and computing power.
The debate surrounding AI investment has become increasingly important as technology companies continue to spend heavily on data centres and computing capacity. Investors are trying to determine whether the expected future revenue from AI will justify today’s infrastructure expenditure.
Supporters of the AI investment cycle argue that the technology is creating new products, improving productivity and opening new markets. Companies across sectors are integrating AI into customer service, software development, research, healthcare, finance and other areas.
Vembu does not appear to reject these possibilities. His position is more cautious: technological potential should not automatically be used to justify unlimited investment.
This distinction is particularly relevant for businesses deciding how much money to commit to AI. A company may benefit from adopting AI without necessarily needing to build the largest possible infrastructure or spend aggressively simply because competitors are doing so.
The comparison with the telecom bubble also provides a historical warning. During the late 1990s, investors were correct about the transformative potential of the internet and telecommunications. What proved incorrect were some of the assumptions about how quickly companies would generate returns from the enormous infrastructure investments being made.
The same pattern could potentially emerge with AI. The technology may continue to improve and become more deeply integrated into everyday life, while the financial expectations surrounding individual companies or projects could prove unrealistic.
Vembu has previously expressed scepticism about some of the more optimistic claims surrounding AI. In May 2026, he described the current AI wave as potentially the biggest technology bubble yet and questioned whether the industry was seeing productivity gains proportional to the scale of investment.
His latest comments therefore fit into a broader pattern of caution rather than representing a sudden change in his position.
He has also questioned claims that AI alone is responsible for job cuts in the technology industry. According to his earlier arguments, companies may sometimes attribute layoffs to AI because it presents restructuring as technological transformation, while financial pressures and other business considerations may also be involved.
The larger question is whether the enormous investment in AI will eventually generate enough economic value to justify the cost. If AI productivity rises substantially, the current spending could prove rational in hindsight. If returns remain below expectations, companies could face pressure to reduce investment.
There is also a difference between investment in productive infrastructure and speculative investment. Data centres, chips and computing systems can have long-term value even if the companies financing them experience difficulties. Infrastructure can remain useful after a particular investment cycle ends.
That was one of the key lessons from the telecom boom. Much of the infrastructure created during the bubble eventually became useful, even though many companies that financed or operated it did not survive the collapse.
A similar outcome is possible in AI. Even if some companies fail, the computing infrastructure built during the current expansion could continue supporting future AI applications.
This could mean that a correction in AI-related financial markets would not necessarily mean the end of artificial intelligence. Instead, it could lead to consolidation, lower valuations and a shift towards projects capable of demonstrating clearer economic returns.
Vembu’s warning is therefore aimed more at financial expectations than at the underlying technology. He remains optimistic that AI will ultimately fulfil its potential, even while cautioning businesses and investors about the risks of excessive enthusiasm.
This distinction is increasingly important as AI becomes part of mainstream technology. There is a temptation to treat every increase in AI spending as evidence of future growth. But spending alone does not prove that the resulting products will generate sufficient revenue or productivity gains.
Companies ultimately need customers, sustainable business models and measurable benefits. If those elements do not develop quickly enough, high infrastructure costs can become a burden.
The impact is already extending beyond AI companies themselves. When major technology firms compete for the same chips, memory and power resources, other industries can face higher costs or longer supply timelines.
Consumer electronics manufacturers are particularly exposed because smartphones, laptops and other devices depend on many of the same semiconductor and memory supply chains. The AI boom can therefore indirectly affect consumers who have never used an AI service.
The rise in memory demand is especially important because modern AI systems require enormous quantities of high-performance memory. As AI data centres expand, suppliers must balance demand from AI infrastructure against demand from traditional computing and consumer electronics.
This creates a complicated supply environment. If manufacturers increase production too aggressively, they risk excess capacity if AI demand slows. If they expand too slowly, shortages and high prices could persist.
Vembu’s concerns therefore extend beyond investment valuations to the structure of the technology supply chain. The AI boom is influencing decisions about manufacturing capacity, power generation and infrastructure planning.
For software companies, the situation could also encourage a renewed focus on efficiency. If computing resources remain expensive, businesses may have greater incentives to reduce unnecessary processing and memory consumption.
The AI industry itself could eventually move in the same direction. Early development has often focused on building increasingly large models and expanding computing capacity. Over time, companies may place greater emphasis on smaller, more efficient models that can provide useful results at lower costs.
Such a shift would not necessarily represent a failure of AI. Instead, it could indicate that the industry is moving from an investment-heavy experimentation phase towards a more commercially disciplined phase.
Vembu’s comparison with the telecom bubble also highlights the difference between infrastructure and business models. Fibre-optic networks eventually became extremely valuable, but not every company that invested in them was able to capture that value.
Similarly, AI infrastructure may become essential while some businesses currently valued highly by investors fail to generate sufficient returns.
The warning is particularly relevant for investors because technological excitement can sometimes make it difficult to distinguish between a promising technology and a promising investment. A technology can succeed while individual companies, projects or investments fail.
For businesses, the lesson is to focus on actual use cases and measurable returns rather than adopting AI simply because it is a major industry trend. Companies need to determine whether AI improves productivity, reduces costs, creates new revenue or enhances their products in ways customers are willing to pay for.
For consumers, the effects of the AI boom may increasingly be visible through hardware prices. If memory and semiconductor demand remain elevated, the cost of smartphones, laptops and other devices could remain under pressure.
The situation also demonstrates how developments in one part of the technology industry can affect completely different markets. AI may appear to be primarily about software and algorithms, but its growth depends heavily on physical infrastructure.
Data centres need land, electricity, cooling systems, networking equipment and specialised hardware. These requirements connect AI investment with energy, construction, semiconductor manufacturing and other industries.
The scale of these requirements explains why investors are closely watching whether AI revenues eventually catch up with infrastructure spending.
Vembu’s warning does not provide a prediction about exactly when a correction might occur. Instead, it highlights the possibility that the current pace of investment may not be sustainable indefinitely.
His central message is that optimism about technology should be accompanied by financial caution. Companies can believe in AI’s long-term potential while still questioning whether every current investment is justified.
That approach also reflects the lessons of previous technology cycles. The internet transformed the global economy despite the dot-com crash. Mobile communications became fundamental despite the telecom bubble. Technologies can survive financial corrections and emerge stronger after excessive expectations are removed.
Artificial intelligence could follow a similar path. A possible correction in AI-related investment would not necessarily stop technological development. It could instead force companies to focus more carefully on efficiency, commercial viability and real-world applications.
The challenge for the industry will be maintaining innovation while avoiding excessive financial risk. Businesses that can demonstrate genuine value from AI may be better positioned if investment conditions become more difficult.
For investors, the coming years could therefore involve a sharper distinction between companies with sustainable AI business models and those whose valuations depend primarily on expectations of future growth.
Vembu’s comments serve as a reminder that technological revolutions and investment bubbles can happen at the same time, but they are not the same thing. AI may ultimately transform industries even if today’s investment boom turns out to have gone too far.
His message is consequently neither an outright rejection of artificial intelligence nor an endorsement of unlimited spending. Instead, he is calling for a more careful approach in which the long-term promise of AI is separated from the short-term excitement surrounding its financial expansion.
As the AI industry continues to attract unprecedented levels of capital, the key question will be whether real-world productivity, revenue and economic value can eventually justify the scale of today’s investment. If they can, the current spending could help create infrastructure for a major technological transformation. If they cannot, the industry may face a correction similar to earlier technology bubbles.
For now, Vembu’s position remains clear: AI technology itself can still achieve its full potential, but investors and companies should not assume that every dollar flowing into the current AI boom will produce a successful return. His comparison with the telecom bubble is ultimately a warning to separate genuine technological progress from financial exuberance.
