Tech IPOs, GenAI & datacenters

The following blog post contains excerpts of a presentation shared during our recent webinar, ‘The Big Picture: Charting the Course for Capital Markets in 2025’. Access the on-demand replay of the webinar and download our Big Picture outlook report for further insights on interest rates, the IPO market, dividends, and other key factors shaping the capital markets in 2025.

I think we are going to have a recovery in tech IPOs, but that seems to be a safe bet. The tech IPO window is usually widely open for one to two years every 10 or 15 years. When this window opens very widely, you usually have a big spike in the number of tech IPOs.

The last widely open window was in 2021. Before that, it was in the 2000s during the dot-com bubble and [in] between, you also had some moderately open windows [with] very robust IPO activity. What we've had over the past three years is just a complete shutdown of the IPO window. The IPO numbers have trended well below pre-pandemic levels.

This is one of the reasons why we think 2025 can be the year where this window probably opens. Not as robustly as it was in 2021 or 2000, but we might have a return to normal, pre-pandemic levels. That would mean a tripling of the number of tech IPOs compared to last year.

And we believe that [because] of the following reasons.

Valuations in the tech space have stabilized. If we look at the closely followed software-as-a-service space, valuations ticked up in 2024. They're still down compared to 2021, but I think this presents an opportunity for a lot of PE and VC investors to exit at a better price than [previous years]. A lot of these private equity and VC investors are facing pressure from their own investors to return capital. So far, I think they found creative ways to avoid getting their companies to market, like continuation funds. But 2025 could be the year where more and more are [going public].

More artificial intelligence [companies and] beneficiaries could also come to market. Reddit had very good performance since its IPO. Reddit is not an AI company, but it benefits from the AI wave. They have hugely valuable data, and they're selling that data to AI companies. Another IPO like Astera Labs, which produces AI chips, also had very good performance since coming to market. I was surprised more AI companies [didn’t come] to market in in 2024, but we could see more in 2025.

The big question, however, remains the path of interest rates, which does impact tech IPOs. Treasury yields have gone up substantially so this represents a big risk to the performance of equity markets. Broadly, we think this is just a bump in the road to the continuation of the bull market.

GenAI funding trends

Generative AI (GenAI) has been one of the bright spots in PE and VC investing. Public and private valuations in this sector have gone up substantially, as you can see from the chart. Prime AI companies have no problems raising billions of dollars. This sector [saw] a doubling of PE/VC investments in 2024 and we expect the fundraising records to probably continue in '25.

Most of the fundraising that we track in the GenAI sector is happening [with] U.S. companies. If you go deeper into the GenAI sector, we see a substantial increase in funding for the companies in the AI infrastructure layer.

AI infrastructure includes hardware, like chip designers, manufacturers, datacenters, software infrastructure, companies that help other companies with data management capabilities, even AI cybersecurity. This [is the] software stack that helps organizations implement AI or prepare organizations for AI adoption. Funding for AI infrastructure companies more than tripled in 2024, and this is not surprising.

If you look at recent surveys of large organizations by 451 Research, spending priorities for AI infrastructure, like the cloud, have continued to grow while [spending on] actual AI technologies has declined a little bit. When [companies] try to implement AI deeper [across] their organizations – not just calling an API from OpenAI or other foundation model providers – they find out that their infrastructure stack is not prepared. So, they go back to basics and upgrade their IT infrastructure and invest in things like data management, data capturing, and further digitize their organizations by adopting all sorts of cloud technologies.

A lot of the value in the AI value chain has been created mostly in the AI infrastructure layer. However, owe're already seeing signs of that it is moving slowly up to the application layer.

Datacenter buildout

Another interesting aspect that we need to talk about is the capital expenditures (CapEx) surge by the hyperscalers. They are expected to continue to grow. Our data suggests that the five hyperscalers will spend collectively $1 trillion between 2024 and 2027, just on CapEx. Most of that CapEx goes to AI datacenters.

We think that number could easily go higher rather than lower given the arms race that the hyperscalers are engaged in. They have deep pockets and most of them have not yet tapped capital markets to fund this expansion. They are using their own cash flows. Only Meta Platforms, for instance, issued a record $10 billion in debt for the specific purpose to invest in new datacenters.

We see datacenters growing [by] double digits, both traditional datacenters and AI datacenters. Some of these datacenters are owned outright by the hyperscalers, but some are rented from specialized providers, so we're expecting strong growth in the leased datacenter market.

If you look at our data, M&A and funding for datacenters is almost equal [as it was] in 2021, with a lot of activity found in the Asia-Pacific region. The valuations of these datacenters are eye-watering. They're trading [like] high-growth software companies. You have 80x-100x earnings valuations across the U.S., Asia-Pacific, and Europe.