AI Growth Now Limited by Computing Power, Not Model Quality: Goldman Sachs

The report highlighted that as demand for AI applications accelerates, particularly in real-world deployment or "inference" use cases, the need for compute infrastructure such as high-performance chips and data centres is growing faster than supply. This has made access to reliable and cost-efficient compute a key differentiator for companies operating in the AI space.

It stated, "Compute has emerged as the binding constraint in scaling AI. As inference demand grows faster than available capacity, the key differentiator is no longer just model quality, but the ability to reliably access and finance compute in the most performant way".

It noted that the focus in the industry is shifting away from just building better models to how effectively these models are run in production environments. Factors such as cost efficiency, reliability, performance, and the ability to route workloads across systems are becoming critical in determining competitive advantage.

According to the report, value in the AI ecosystem is increasingly moving towards the execution layer, including technologies that manage model deployment, optimise compute usage, and ensure consistent performance. Companies that can secure and efficiently utilise compute resources are better positioned to scale their AI offerings.

The report also pointed to a broader structural shift in the software industry. AI-native companies are not directly competing with traditional software-as-a-service (SaaS) firms but are instead targeting gaps between existing solutions. While SaaS platforms have historically focused on systems of record and functional silos, AI-native firms are building systems of action that deliver end-to-end outcomes.

These AI-driven solutions are being deployed faster, often going live within weeks, and are seeing higher conversion rates from pilot projects. Unlike traditional models that charge based on user seats or features, AI-native companies are increasingly monetising based on business impact and productivity gains.

Goldman Sachs further noted that applied AI is approaching an inflection point similar to the generative AI boom seen in 2022. While large-scale commercialisation may still take 5-10 years, advancements in foundation models, simulation, and edge technologies are enabling AI systems to move beyond decision support into real-world execution.

This includes applications across logistics, labour automation, and defence, where AI systems can act, learn, and improve continuously. Companies with sustained real-world deployments are expected to gain an advantage by building proprietary data ecosystems that enhance performance over time.

So the report outlined that as AI continues to evolve, control over compute resources and the ability to deploy models efficiently will play a central role in shaping the next phase of industry growth.

- ANI