'AI bubble might be deflating': Zoho's Sridhar Vembu says real engineering work still needed

The AI gold rush is losing its shine, and Zoho founder Sridhar Vembu isn’t surprised. On Monday, Vembu took to social media to caution against the inflated expectations surrounding artificial intelligence, noting that many corporate customers and analysts remain skeptical. 

"One fact about the recent AI hype is that corporate customers and analysts are not terribly excited," Vembu stated. "They are in a 'prove it to me' mode, as they should be. I am in the camp that overhyping anything is a bad idea. I am personally enthusiastic about some technologies, but I will not overhype them."

He pointed to two recent developments as indicators that the AI bubble might be deflating. "There are some warning signs of the AI bubble deflating: 'Microsoft reportedly cancelled US data center leases amid oversupply concerns'. Satya Nadella cautioned against overly optimistic projections about AI, in particular about AGI (AI that is superhuman in every dimension)," Vembu wrote.

Despite the cooling excitement, Vembu acknowledged AI’s genuine business applications, listing areas where the technology is already making an impact. He said that AI is useful for speech-to-text, text-to-speech, and image recognition for authentication; image-to-text conversion for extracting metadata from images and PDFs; spam, phishing, and fraud detection; security threat analysis; pattern recognition in business, financial, legal, medical, and engineering data; software development assistance, particularly in code generation and debugging; and marketing content creation for websites, brochures, and campaigns. 

Vembu warned that misuse of AI-generated marketing materials could lead to "AI slop," where consumers instantly recognise machine-generated content. He stressed that many of these AI applications predate large language models (LLMs). "We used to call this field 'machine learning,' and those tools are in widespread use," he noted.

Among the listed categories, Vembu said he was most excited about security analysis, pattern recognition, and programmer assistance, particularly autonomous AI agents. However, he cautioned that such agents are "vastly overhyped" and likened their challenges to self-driving cars. "In that sense, autonomous AI agents are like self-driving cars: don’t get too taken by the hype, but engineers in the trenches are making progress," he wrote.

Looking ahead, Vembu said that AI’s true potential will require extensive research and development — not just financial investment. "We are going to need a lot more R&D work, and that does not mean throwing a lot more money at it. When hype meets Wall Street, a lot of money gets thrown at it (be it the optical bubble or the real estate bubble and now the AI bubble), and then Wall Street gets tired and walks away. That is when the real engineers get excited."