Why the B2A era could be the next growth engine for AI

The AI shock from “Pacing the Frontier”

Calls to pace the development of frontier AI models earlier in September led to an initial selloff in global semiconductor stocks by around 6%. By 21 September, losses were erased as the sector had rebounded by around 11%. The success of consumer AI applications, such as Meta’s Muse app, and agentic capabilities quickly brought attention back to AI adoption and growth. The reversal illustrates how quickly concerns about the investment cycle can give way to renewed confidence in the opportunities offered by AI.

Proposals to “pace” the frontier AI models largely centre around advancing AI responsibility, rather than stopping model development or reducing AI spending altogether. Furthermore, competition for technological leadership, reinforced by the strategic ambitions of the US and China, makes a coordinated slowdown challenging. In our view, investors should continue to look past near-term volatility and focus instead on underlying fundamentals.

That said, the correction still reflects a key concern among investors: whether the scale of AI investment is supported by sustainable demand. Meanwhile, our industry enquiries suggest that the combined revenue backlog of three leading cloud platforms has reached around USD 1.7 trillion, increasing by more than USD 1 trillion over the past year. Accelerating cloud revenue growth and solid return on AI investments from hyperscalers further support the case for their continued infrastructure commitments.

Entering the B2A era

While much of the recent debate has focused on “Pacing the Frontier” and the sustainability of AI infrastructure spending, the success of personal AI assistants such as Meta’s Muse highlights a new source of growth through the “agentification” of the consumer AI trend. For example, with only a few prompts, consumer AI agents can seamlessly identify the best pricing options across airline tickets, hotels and telecom packages, while autonomously completing transactions on behalf of users.

The opportunity for platforms providing these assistants goes beyond attracting new users. As agents become more capable of completing online purchases, platforms could earn fees on successful transactions, creating another channel for monetisation. We expect leading AI labs to follow suit, expanding their opportunity beyond enterprise customers into a potentially larger consumer market over time.

We see this as the next stage of the digital economy – from business-to-business (B2B) and business-to-consumer (B2C) towards the business-to-agent (B2A) era.

However, these benefits are unlikely to be evenly distributed. Online travel agencies, marketplaces and price-comparison services that fail to adapt, or adapt fast enough, face the risk of disruption. Businesses will thus need to offer services that are accessible to these agents, similar to how they adapted their websites for search engines.

What does this mean for investors?

If AI agents become the primary interface for consumers to search, evaluate and transact, the resulting increase in inference workloads could drive significant expansion in compute demand. Serving billions of agents would require more processing capacity, memory and storage, while central processing units which orchestrate agent workflows appear among the clearest near-term beneficiaries. The opportunity for the semiconductor industry therefore extends well beyond the race to train the most capable model.

Against this backdrop, we remain constructive on the semiconductor sector. We are also positive on Big Tech, where expanding demand for compute should support both cloud growth and new AI monetisation opportunities.

Beyond technology, financial services, healthcare and consumer sectors that successfully integrate AI into their business models are also likely to be well positioned to benefit from the emerging B2A economy.