AI could outperform 90% of coders in six months: Zoho's Sridhar Vembu

In this article, “outperform” means an AI system can create functional computer code more quickly and efficiently than a human software engineer. It describes practical performance, not a claim that AI has replaced every part of engineering work. The comparison matters because coding has traditionally depended on skilled human labor.

Vembu compares today’s change with powerful machine looms replacing handloom work. Automated systems can generate complex scripts at high speed. He says AI already writes better code than at least 60–70% of software engineers. The key mechanism is software automation: machines produce working code while reducing the time and effort needed from people.

This could make traditional coding less valuable and change the role of many developers. It does not mean every engineer disappears immediately. Zoho’s response shows another possibility: redeploying engineers to build server infrastructure and other deeptech hardware. The article presents AI performance as a major industry shift, while leaving the exact definition of “better” broad.

Vembu describes AI coding improvement as extremely rapid. He says current systems can already write better software than at least 60–70% of software engineers. His expectation is that progress will accelerate further over the next three to six months, allowing AI to outperform about 90% of engineers.

The mechanism is faster automated software creation. AI systems can generate functional code and complex scripts at high speed. As these systems improve, they may handle more coding tasks with less human effort. Vembu compares this shift with powerful machine looms emerging in a former handloom industry.

The forecast points to intense pressure on traditional coding jobs. However, the article presents it as Vembu’s prediction, not a measured future result. Zoho is preparing by moving engineers into deeptech hardware and infrastructure work. The broader challenge, he argues, is deciding whether AI increases national technology ownership or deepens dependence on foreign systems.

If AI creates software much more cheaply, companies may need fewer people for traditional coding work. The article describes this as a major disruption for technology professionals. It also says lower software creation costs could spread beyond the technology sector, because automated efficiency may reduce the expenses involved in producing goods and services.

The key mechanism is deflationary pressure. When production becomes cheaper, consumer goods may cost less. Teachers, nurses, barbers, and other non-technology workers could then gain more real purchasing power, even if their jobs are not directly connected to software. Vembu says roughly 97% of the wider population does not write software and mainly consumes technology.

The effects would not be evenly distributed. Major technology hubs could face direct upheaval if coding jobs disappear, while places such as Tenkasi and Patna may see little immediate impact. The article frames cheaper production as a possible broad benefit, but it also stresses that software workers face serious disruption.

Vembu distinguishes software labor from software ownership because the two produce different economic returns. Indian engineers may write or design important technology, yet foreign companies can retain the intellectual property and capture the main financial gains. For him, the central national problem is not losing outsourcing jobs alone, but lacking ownership of the technologies being created.

He points to Indian engineers writing much software for Microsoft and Cisco, and designing 40–50% of Nvidia’s chip design work. Yet the primary returns flow to foreign intellectual property holders. His comparison is stark: a Swiss teacher can earn an iPhone in three or four days, while an Indian teacher may need months.

The implication is that services exports can reward labor without building technological sovereignty. Vembu argues India should develop and own foundational layers, as East Asian economies did. Ownership would shape the income of non-technology workers and reduce dependence on foreign intellectual property.

Zoho is responding to AI disruption by keeping its engineers and moving them into different technical work. Vembu says the company does not lay off employees because it believes they can be redeployed. This approach treats AI as a reason to redirect engineering capacity rather than simply remove it.

The immediate focus is deeptech hardware, including building server infrastructure in-house. The move follows a sharp rise in infrastructure costs linked to AI adoption. The article says server prices climbed from $8,000 to $45,000, while graphics processing unit costs also exploded. Internal construction could help Zoho address these demands directly.

This strategy reflects Vembu’s broader argument that foundational technology matters more than rushing toward top-level AI applications. Zoho’s engineers may face pressure from current AI systems, but the company sees useful work in hardware and infrastructure. The response combines workforce redeployment with an effort to strengthen technological capability inside the company.

Vembu compares India’s software-services model with East Asian development paths because he believes labor-based outsourcing does not create enough national technological independence. India became known for writing software for foreign companies, but the ownership and financial returns often remained elsewhere. He says Western commentary in the 1990s encouraged this focus.

His preferred model builds domestic technology layers before chasing advanced applications. China, for example, developed national platforms such as Baidu for search and WeChat for messaging. Japan, South Korea, Taiwan, and China are presented as countries that pursued stronger technology foundations and ownership. The mechanism is control over the systems that support later innovation.

Vembu argues India should apply a similar strategy instead of obsessing first over AI. Foundational capabilities could include hardware, operating systems, lasers, X-rays, robotics, and agricultural technology. The intended result is lasting economic independence, rather than continued reliance on foreign intellectual property, hardware, and operating systems.

Vembu says countries should build foundational technologies because national prosperity depends on owning the systems beneath advanced applications. These include hardware, operating systems, robotics, lasers, X-rays, and agricultural technology. Control of such layers can support wider economic independence and determine who captures technology’s value.

The mechanism is dependence. A country may develop advanced AI software, but if it relies entirely on foreign hardware and foreign operating systems, outside suppliers still control essential infrastructure. Vembu warns that this dependence creates extreme geopolitical vulnerability. His concern extends beyond software jobs to the physical and technical foundations that make modern systems operate.

That is why he says India should not obsess about AI first. Zoho is building server infrastructure internally and redeploying engineers toward deeptech hardware. The article presents this as part of a broader shift from exporting services to securing genuine national technology ownership. Stronger foundations could make future AI development more resilient and self-reliant.