Cheaper AI models could change which chips and infrastructure companies need, but the available statements from Infosys CEO Salil Parekh do not establish that lower model costs are reducing data-centre capacity usage. On Infosys’s July 2026 Q1 FY27 media call, he discussed smaller models, token costs and memory-chip expense—not a proven decline in data-centre utilization.
What did Infosys CEO Salil Parekh say?
Parekh said Infosys’s smaller language models for banking and IT operations were gaining traction. He described their smaller parameter sets, lower running costs and more efficient token use. His remarks concern Infosys’s offerings; they are not an independent comparison of model costs or a forecast of industry-wide infrastructure demand.
As an Amazon Associate I earn from qualifying purchases.
Discussing the client environment, Parekh said, “We see, for example, today, there’s a tremendous attention to token cost.” He also pointed to memory-chip expense and the ongoing need for efficiency. These comments suggest buyers are scrutinizing the cost of using AI and the infrastructure behind it, but do not say that customers are using less data-centre capacity.
The MarketBeat transcript of Infosys’s July 2026 Q1 FY27 media call records those comments. The exact claim that high data-centre and chip costs, combined with cheaper models, pose a risk to capacity usage is not verified by the available transcript.
How could cheaper models affect infrastructure demand?
A smaller or more efficient model may lower the compute needed to complete a given task. That can reduce the cost of inference, particularly if the model can handle work that otherwise would have gone to a larger system. Infosys’s 2026 analysis frames the economics in terms of cost per task, workflow or autonomous loop, rather than cost per token alone. It also says the business effects of recent model-size innovations remain uncertain.
Lower cost per task does not automatically mean lower total infrastructure use. If cheaper inference makes AI affordable for more users or more tasks, overall workloads may grow. Demand can also shift among types of processors and memory rather than disappear. The available material does not provide a like-for-like benchmark that shows whether savings from smaller models outweigh any growth in use.
For buyers, the practical question is therefore not simply whether a model is smaller. It is whether the full workload can run at the required cost, speed and scale while meeting software, integration, data-governance and supply needs.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What do Infosys’s chip forecasts show?
Infosys Knowledge Institute’s Semiconductor Industry Outlook 2026 describes expanding AI server and memory demand, alongside rising costs in parts of the chip supply chain. Its figures point to a changing mix of infrastructure demand, not evidence that installed data-centre capacity is going unused.
| Measure | What the outlook says | How to read it |
|---|---|---|
| AI server shipments | Cloud service providers’ ASIC-based AI server shipments are projected to grow approximately 45% in 2026, compared with 16% growth projected for GPU-based AI server shipments. | These are Infosys projections, not final shipment results or utilization measurements. |
| DRAM revenue | The report says DRAM revenue tripled from $50 billion in 2023 to more than $150 billion in 2025. | This is the report’s historical estimate, attributed to Infosys. |
| High-bandwidth memory (HBM) | HBM’s share of DRAM is projected to rise from 5% in 2023 to 50% in 2028. | This is a projection, not a reported outcome for 2028. |
The projected difference between ASIC- and GPU-based server shipment growth indicates that demand may shift across hardware categories. It does not show that one type is interchangeable with the other for a particular workload, nor does it reveal how much of the installed fleet is being used. The outlook is an Infosys Knowledge Institute analysis, not an independent utilization survey. Read Infosys Knowledge Institute’s Semiconductor Industry Outlook 2026.
Does the evidence show a capacity-usage risk?
No. The available statements connect AI adoption with attention to token costs, memory-chip expense, smaller models and efficiency. Infosys’s semiconductor outlook discusses server shipment projections and memory demand. None of those facts establishes that high infrastructure costs or cheaper models are causing data centres to run below capacity.
Utilization can also mean different things. In Infosys’s Q4 FY26 press-call discussion, the CFO linked staffing partly to business volume and utilization. That concerns workforce and company operations, not the utilization rate of data-centre equipment. Infosys’s SEC-filed transcript also discusses AI-services opportunities, competitive intensity and productivity impacts, but does not establish a data-centre capacity trend. See Infosys’s Q4 FY26 press-call transcript filed with the SEC.
What would confirm the headline’s concern?
To establish a genuine capacity-usage risk, reporting would need evidence that links the cost or model shift to actual infrastructure outcomes—for example, measured utilization, delayed or cancelled capacity plans, or workloads moving away from existing facilities. The remarks and forecasts cited here do not supply those measurements. Until the originating report or interview behind the headline is available, its causal framing should be treated as unverified, not as a confirmed statement by Parekh.
Best Value
- Family farms not data design for people against AI server farms, data center expansion, rural land buyouts, corporate agriculture, and industrial tech development replacing farmland and open space. Rural conservation and anti data center message.
- AI protest design for farmers, land conservation supporters, anti AI activists, sustainability groups, environmental advocates, rural communities, and people opposing server farm construction, power grid strain, and farmland destruction.
- Lightweight, Classic fit, Double-needle sleeve and bottom hem
For now, the defensible conclusion is narrower: model efficiency and infrastructure costs are important to buyers, and AI server demand may evolve toward a different hardware mix. Whether that creates stranded or underused capacity remains an open question.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
