“I guess you guys aren’t ready for that yet. But your kids are gonna love it.” AI hype seems to be channeling Marty McFly. USD 40bn of enterprise investment in generative AI has delivered scarcely any return to corporates, yet there is an undeniable earthquake in technological possibilities.

Adding to the market’s déjà vu, an increasingly clear parallel between the AI run and the dotcom bubble of the late 1990s is raising the specter of stock market fragility.

Dealmaker whispers about the sustainability of AI-focused valuations, driving wider market sentiment, have been growing in volume these past weeks.

Even though the TACO trade seems back on following last Friday’s wobble, the NASDAQ-100 took a 3.49% knock while chips behemoth NVIDIA, the best listed play for AI hopes, fell 4.9%.

Mega-cap tech stocks – embodied by the Magnificent Seven – are pricing in perfection and seem ripe for profit taking, or worse. The valuations of AI giants pressing the cause “appear stretched”, the Bank of England’s Financial Policy Committee said last week.

It’s tough to disagree.

The Magnificent Seven account for 32% of the S&P 500’s aggregate market capitalization but just 23% of forward earnings, according to Yardeni Research.

Had the NASDAQ Composite continued its trend growth from the post-Lehman period through to 2020 – rather than rocketing higher amid COVID-19 and AI-related bull runs – we’d expect the index to be sitting closer to 16,000 than its current level just above 22,000. That would imply a 27% fall; in April 2000, the index sank 25% in one week as the dotcom bubble burst.

Indeed, the scale of AI investment since 2023 broadly matches the capital put to work rolling out the internet from 1995-2000 in GDP percentage terms.

Then, as now, a generationally defining technology promises to upend corporate workflows and have major societal ramifications. Then, as now, irrational exuberance risks allocating capital too freely and with insufficient certainty on who the winners are likely to be.

At present, the major AI arms race is focused on building capacity, namely the processing power and data center networks needed to drive the revolution over the coming decade. Some USD 7tn in investment is required to meet the demand for compute power by 2030, McKinsey has estimated.

Two strategic partnerships unveiled this past month by OpenAI with chips giants NVIDIA and AMD threw this into sharp focus. Both involved eye-watering data center gigawatt commitments and stock purchases. Meanwhile, OpenAI investor Softbank is reportedly exploring a USD 5bn margin loan backed by Arm stock to finance AI capex.

This all sounds disconcertingly circular – and attempts at financial alchemy should be a warning sign even if you have conviction on AI’s long-term fundamentals.

NVIDIA’s 3Q numbers on 19 November will be a critical catalyst. Any hint of slowing growth in demand, rollout, capex execution or technological development would dent sentiment in a febrile market.

But continued robust performance would buy further time for AI investments to deliver tangible gains in the wider corporate landscape.

There may be tentative signs of these coming through. Minutes from the Federal Open Market Committee (FOMC) meeting last month revealed some members had argued that efficiency gains could be moderating inflation pressures.

Alongside the impact of changes in US immigration policy, increased productivity driven by AI would offer a benign read on softening jobs numbers, from a GDP perspective.

This is a pretty rosy picture and one not truly evident from the data so far. Though enormous numbers of dollars are being put to work building AI capacity, many American corporates are stuck wondering how to invest.

One New York-based mid-market sponsor partner with a tech focus told Continental Drift a CTO at one of his portfolio companies won’t move forward with hiring graduates as he can’t guarantee they’ll have a job in six months’ time, so fast moving are technological capabilities.

AI is “more impactful in theory than in practice so far”, the partner added. “We’re not getting the benefits – yet. Once you do, you won’t need that side of the building.”

That “yet” is at the heart of the matter.

Corporates are uncertain what to back in this fast-paced environment, for fear of succumbing to first mover disadvantage. To take one example, does KKR’s longstanding position in WebMD owner Internet Brands, valued at USD 12bn in 2022, represent a solid investment in content and data vital for LLM training models, or is it an information Kodak in a GenAI world of digital clarity?

Even if the latter is closer to the truth, the jury’s still out as to whether the hyperscalers themselves reap all the benefits. They may do the rollout work only to see more nimble service users cream off the benefits.

Amazon is in the Magnificent Seven; Verizon and A&T are not.

Corporate focus on lean operations, mid-market sponsors’ need to generate more strategic interest in long-held assets, and building readiness for the technological tsunami coming our way are evergreen themes for US dealmakers to pursue as they execute on strong year-end M&A pipelines.

“Where we’re going, we don’t need roads.” Just as well if some air’s coming out of the tires. Prepare for a correction before you run “OUTATIME”.

Continental Drift is a weekly column offering commentary on the macroeconomic, political, and policy forces shaping the M&A landscape across the US and Europe. The opinions expressed here are those of the writer only.