Truth that Matters. Stories that Impact

Truth that Matters. Stories that Impact

Business

Sridhar Vembu Warns Surging Memory and AI Token Costs Are Straining Tech Businesses

Zoho founder Sridhar Vembu has cautioned that doing business in the technology sector is becoming increasingly challenging as memory prices surge by up to 500 percent over a 12-month period alongside climbing artificial intelligence token costs.

What Happened

Sharing his concerns on social media platform X, Sridhar Vembu cited industry data showing that memory prices have jumped 500 percent over the past year. Vembu highlighted that the combination of steep memory price hikes and elevated AI token expenses has put severe pressure on operational costs. While Zoho has resisted passing these costs on to customers through price increases so far, Vembu stated that holding off is becoming hard.

The sharp increase in hardware costs is illustrated by data from a Tom’s Hardware report. Between August 2025 and August 2026, average prices for DDR5 memory rose from $90 (approximately Rs 8,600) to $425 (approximately Rs 40,000) for certain kits. High-capacity variants witnessed even steeper jumps; for instance, a 128GB DDR5-6400 kit reached $3,399 (roughly Rs 3,25,000)—ten times its previous low of $329 (roughly Rs 31,490).

Key Highlights

  • Zoho Feels Margin Pressure: Sridhar Vembu noted that rising memory prices and AI token costs are complicating business operations, making it difficult to maintain existing software pricing without increases.
  • Severe Hardware Price Inflation: Certain DDR5 memory kits have surged from $90 to $425 between August 2025 and August 2026, with top-tier 128GB kits experiencing a tenfold increase.
  • Broad Hardware Impact: Smartphone manufacturers including Oppo, OnePlus, Vivo, and Nothing have raised their handset prices in recent months, with rumors that Apple may consider price hikes as well.
  • Call for Programming Evolution: Vembu stated that the assumption that memory is free has ended, arguing for more memory-efficient programming languages and smarter compilers.
  • Heavy Data Centre Investments: OpenAI has partnered with SoftBank’s SB Energy for a 10-gigawatt data centre in Ohio with backing from Nvidia, while Indian firms HCLTech and TCS are committing Rs 3,500 crore and up to $7 billion respectively toward data centre infrastructure.

Why This Matters

The relentless demand for memory chips generated by AI data centres is having a cascading effect across both hardware manufacturing and software development. While device makers such as Vivo, Oppo, OnePlus, and Nothing have already adjusted their retail prices, enterprise software companies that run substantial server workloads are now grappling with cost surges. Vembu noted that code safety and developer productivity must not result in memory bloat, pointing out that programming languages can no longer rely on the premise that memory resources are effectively inexpensive.

What to Watch Next

Vembu indicated that exploring how programming languages can function using less memory forms the core of his current research. In the broader market, observers are monitoring whether enterprise software providers will eventually hike consumer and business prices. Meanwhile, major infrastructure expansion continues, led by massive investments including the 10-gigawatt Ohio facility linked to OpenAI and SoftBank, along with Indian data centre projects initiated by TCS and HCLTech.

Frequently Asked Questions

Why are memory chip prices rising so quickly?

Surging demand for memory chips from global AI data centres has driven a memory supply crunch, pushing certain kit prices up significantly over a 12-month period.

How are tech companies reacting to higher hardware expenses?

Smartphone makers such as OnePlus, Oppo, Vivo, and Nothing have increased device prices, while enterprise software providers like Zoho are finding it challenging to maintain existing rates without raising prices.

What solution did Sridhar Vembu propose for software development?

Vembu stated that developers need to create highly memory-efficient languages and smarter compilers—even for AI workloads—moving away from the longstanding notion that memory is an abundant, low-cost resource.

Source: Sridhar Vembu on X, with reporting from India Today and Tom’s Hardware.