Here’s what’s on our plate today:

• 📝 AI investment isn’t slowing; it’s shifting.

• 🧠 Brain Snack for Builders: Infrastructure doesn’t bubble.

• 🧵 Voice clones, LatAm data bets, chatbot therapy.

• 🗳️ Poll: What’s really inflating this so-called AI bubble?

Let’s dive in. No floaties needed…

The Laboratory

Why Big Tech isn’t afraid of an AI bubble

Historically, before an innovative piece of technology makes its way into the hands of the masses, it undergoes rigorous scrutiny to assess its commercial viability as well as effectiveness. This process often takes a long time, which is invested in developing the infrastructure, proving capabilities, and rolling out adoption.

The process can only be completed with participation from enterprises, big businesses, governments, and finally, end-users whose adoption ensures long-term sustainability.

However, during the process, when the viability of a technology is yet to be proven, developers cannot depend on end-users to bear the cost of development. This, in turn, puts the onus of betting on unproven technological developments on investors, enterprises, and government institutions.

Often, technologies are able to prove their commercial viability, and early adopters and investors gain the most benefit. However, not every technological advancement has the potential to have a meaningful impact.

The Google Glasses is a fine example.  Launched with major hype in 2014, they were positioned to bring AR wearables into everyday life. However, high costs, privacy concerns, usability issues, and unclear value for many users led to its consumer retreat and eventual discontinuation (for the mass market) in 2023.

In 2025, Artificial Intelligence is the technology that promises high returns, prompting huge investments in its development and deployment. However, since the returns are yet to materialize, many are questioning the continued inflow of finances in companies that make AI possible, even going to the extent of calling this investment a bubble.

A slowdown that’s more surface than substance

Over the past few weeks, media coverage has intensified the debate on whether the investments in AI constitute a bubble, warning that a collapse could ripple through the broader economy. Reports highlighted the self-reinforcing loop of AI investment, rising layoffs, and companies struggling to explain the real purpose or payoff of their AI initiatives.

Even leading figures in the field admitted that the technology had not advanced as dramatically as its most enthusiastic supporters had claimed.

A recent report from Business Insider states that in a recent investor update, analysts at RBC Capital Markets suggested that enterprise AI adoption may be slowing, even though major tech companies continue to post strong AI-led performance.

Analyst Rishi Jaluria and his team noted that the robust demand highlighted by Microsoft, Amazon, Meta, Oracle, and Google is largely tied to model training, infrastructure build-out, and spending by AI-first companies, rather than a broad wave of uptake among conventional enterprises.

They cited fresh figures from Ramp’s Fall 2025 Business Spending Report, which show that the percentage of U.S. businesses purchasing AI services dipped from 44.5% in August to 43.8% in September.

Though the decline is small, it represents the first noticeable step back since the enterprise AI boom began in 2023.

Fragmented gains, fading pilots

According to the report, analysts say there could be three reasons for the slowdown. These include the hesitation of enterprises that are still waiting for meaningful productivity gains from AI. They say that isolated improvements may not be enough to push adoption.

According to the report, many tools fail to translate into system-wide efficiency gains, and some organisations retreat into pilot fatigue amid privacy and data-governance worries.

Meanwhile, outside of clear wins in areas like coding, customer support, and marketing, most industries, including healthcare and supply chain, have yet to discover a true breakthrough application.

The productivity mirage

A report from McKinsey echoes similar trends. According to an online survey of 1,993 participants in 105 nations, the current AI landscape can be defined by immense promise but uneven progress.

On one hand, Big Tech platforms continue to report surging demand for AI infrastructure and model development, driven largely by AI-native companies and large-scale training efforts.

On the other hand, a growing body of evidence suggests that traditional enterprises are not experiencing broad, transformative gains.

Additionally, spending data shows the first slight decline in the share of U.S. businesses paying for AI services, signaling a pause in adoption momentum after two years of rapid acceleration. Analysts argue that the most powerful advances have been concentrated in model development, not in everyday business operations.

The report further highlights that AI’s impact remains fragmented, with productivity gains being isolated because companies have not redesigned their workflows or coordinated processes to allow AI to work end-to-end.

Some firms are pulling back on pilots, worn out by inflated expectations and wary of privacy, governance, and data-security risks. Even as agentic AI tools spread across IT, knowledge management, customer support, and marketing, the killer apps that truly transform entire sectors have yet to materialise in areas like healthcare or supply chain.

Why hyperscalers are still spending

So, when AI adoption is witnessing a dip, why are big tech companies continuing to invest more in AI infrastructure?

The points raised by McKinsey and RBC Capital Markets are indicative of how AI is currently stuck in a phase where it is yet to be able to bring about an overhaul of enterprise processes. However, this is not the opinion that dominates the debate around the value of AI.

Economist Martha Gimbel told MIT Tech Review that it would be historically shocking if a technology had had an impact as quickly as people thought that this one was going to. What this means is that much like end-users, most of the economy is still figuring out what AI can do, not deciding whether to abandon it.

Overall, the narrative that emerges is one of a technology advancing at extraordinary speed while the systems meant to absorb it, organisational structures, regulations, culture, and economic incentives, move far more slowly.

This mismatch creates the impression of an AI boom from the outside, while inside enterprise environments, the path to real, scalable impact is proving far more complex.

This may explain why Nvidia CEO Jensen Huang shared that the semiconductor giant is experiencing "very strong demand" for its state-of-the-art Blackwell chips.  Microsoft plans to spend $80 billion on AI‑enabled data centers, and Meta has earmarked $60 billion to build out AI capacity.

Another marker is that, despite talk of unrealistic valuations and unjustified investment in AI stocks, brokerage firms like Robinhood want to give amateur investors a route to put money into private artificial intelligence companies.

Infrastructure doesn’t bubble

Historically, the companies that build and maintain the foundational rails of new technologies are insulated from short-term fluctuations because their role is infrastructural, not dependent on immediate consumer validation. The steam engine took decades to reshape manufacturing; electricity transformed cities long before it transformed office productivity; the internet went through multiple boom-and-bust cycles before emerging as a fundamental layer of global commerce.

In each case, the early builders were sustained not by fads but by the inevitability of scaled adoption. AI sits squarely in that lineage.

The core of hyperscaler confidence rests on a simple but powerful reality: once enterprises begin the shift toward AI-driven processes, the infrastructure needed to support that shift becomes indispensable.

Whether a company deploys one AI agent or a thousand, it must invest in compute, storage, networking, and integration tools. That spending may rise or fall at the margins, but it does not reverse. The hyperscalers feel this every day in their order books, where commitments for next-generation chips, expanded data-center capacity, and long-term cloud contracts continue to grow, even if individual companies slow their experimentation.

Moreover, the current plateau in enterprise adoption is not a sign of retreat but of recalibration.

Early pilots produced excitement but also exposed gaps in data readiness, workforce capability, compliance frameworks, and operational design.

Companies realized that plugging AI into existing workflows yields limited value. To unlock transformative gains, they must redesign those workflows entirely. This redesign takes time, money, and leadership alignment.

From the hyperscaler perspective, these delays are not failures but necessary preludes to scale. Once these foundations are rebuilt, demand for AI services training, deployment, and inference typically accelerates rather than declines.

Are expectations the real bubble?

Crucially, the idea of an AI bubble often misunderstands where the real value lies. The bubble, if it exists, is more in expectations not infrastructure.

And so the story circles back to where it began: society is once again in the midst of a technology’s developmental phase, one that demands patience, sustained investment, and a willingness to endure false starts.

So, while users should be mindful of chatter of an AI bubble, it would be wise to understand exactly where the bubble lies. Whether it is in expectations, in AI’s ability to provide meaningful gains in productivity, in startups that are slapping AI on every piece of automation tech, or in the underlying infrastructure.

In the meantime, it should be remembered that every bubble bursts; however, when they do, much like the dot-com bubble, they leave behind the pipes the next boom will run on.

Brain Snack (for builders)

Quick Bits, No Fluff

• Your voice, forever: Eternos pivots to a personal AI that sounds exactly like you.

• LatAm gets louder: Data center giants pour billions into Latin America’s AI boom.

• Chatbot vs. crisis: A new mental health bot aims to detect eating disorders early.

Thursday Poll

Meme Of The Day

Rate This Edition