Genuine potential of artificial intelligence

The genuine potential of artificial intelligence

Following a recent series of trips to meet leading technology investors in the US, Asia and Europe, we consider the outlook for artificial intelligence and software companies in general.

Executive summary

  • Technology investors and enterprises alike have learned lessons about capital discipline after their experiences from the market euphoria of 2020-21 and their struggles in 2022.
  • Yet throughout these different market regimes, software in particular has continued to be resilient and to show its path towards further strong growth, as software efficiencies can increase corporate productivity.
  • AI has the potential to be a major catalyst, unlocking new efficiencies, and creating attractive investment opportunities across companies that are pioneers, providers, and adopters.
  • However, the long-term victors in such paradigm shifts rarely become apparent in the very short term, so some diversification of exposure is important.

Open AI and Shut AI

A lot can happen in a year – although plus ça change, plus c’est la même chose. OpenAI launched ChatGPT, the first popular large language model (LLM) that sparked mainstream excitement about the near-term potential of artificial intelligence (AI), in November 2022. In November 2023, OpenAI’s board very publicly ousted its co-founder and chief executive – only to reinstate him several days later.

Media reports suggested the reason for his brief removal was the then-board’s discomfort at the company moving too quickly to commercialise its technology. Indeed, its operating agreement explicitly warned investors that “it would be wise to view any investment in OpenAI Global, LLC in the spirit of a donation”.

This evidently did not deter investors from OpenAI or AI as a theme in general. There are many ways to depict the financial excitement around AI over the past year, but perhaps the clearest and most incontrovertible is the surge in revenues at NVIDIA, regarded as the pre-eminent provider of the processors integral to AI’s mass adoption (Figure 1). It is rare to see such step changes in companies already operating at a large scale – the vertical axis is in billions, not millions.

Figure 1
NVIDIA quarterly revenues

Source: Company filings, as of 26 April 2024.

Remember, these real sales – not projections – represent investments only in one key enabling technology for AI; they highlight not simply optimism, but massive capital allocations across the economy in favour of AI’s potential. And these chips are only part of the AI ecosystem and this chart illustrates only one public stock – there is a much broader and deeper pool of opportunities in private markets.

So is the excitement justified?


In conversations with technology investors during our recent trips to the US, Asia and Europe, the levels of AI excitement varied from high to extremely high.

Some investors see it as a category in its own right, and will dedicate funds and assets to it, whereas others view it more as a catalyst for almost every category (from areas of technology like enterprise software, cybersecurity and fintech to consumer-facing apps and services and healthcare for enhanced diagnostics, discovery, or support services).

Many of these investors were active in AI long before ChatGPT became a household name. What has ignited them more recently is the unprecedented speed of creation for new applications; just as the internet was well known before mobile connectivity and the cloud supercharged the app economy, AI may now be going through a similar transition from important to essential, from available to ubiquitous.

AI may now be going through a transition from important to essential, from available to ubiquitous.
— Stanislas Chanavat, Principal Thematics - Private Equity, Pictet Alternative Advisors

One of the critical breakthroughs in facilitating this has been LLMs, which understand human language. This makes them intuitive to use, most obviously in chatbot settings, helping them penetrate the popular consciousness. But perhaps as significantly economically in the nearer term, they can also code. This gives many industries a tremendously powerful and cost-effective new tool, as an enabler of higher productivity rather than replacer of labour because AI’s adoption and real business applications will not be implemented overnight. By the time it can take over human work more fully, we expect it – like every other transformative technology in history, from the railroad to the internet – to have created new and better jobs.

Back to the future

But what gives us confidence in AI as an investment opportunity, rather than just as a technology?

Part of the answer is that, in speaking to the top-tier managers with whom we invest, the industry has learned lessons from its experiences in the past few years. Among the most important are that rigorous due diligence matters more than chasing hot ‘blitzscalers’ and that trends shouldn’t be extrapolated too aggressively.

These lessons are today being put into practice by investors and technology companies. Cash burn rates are being reduced significantly, and the cash runway is in focus. This is evident in employment levels across the US tech sector, for example. Companies from start-ups to giants reduced their salary costs as they came to terms with the new economic environment, with more than 400,000 roles cut since the start of 2022 (Figure 2).

Figure 2
Monthly and cumulative US tech redundancies

Source: TechCrunch and, as of 26 April 2024.

What we also see, however, is that the pace of layoffs has moderated substantially since late 2022 and early 2023. Indeed, we heard in our conversations that recruitment is starting again – but at salaries that are 20-50% lower for the same roles than at the market peak. These rationalisations put tech companies on a surer pathway to profitability.

At the same time as this discipline is taking root in tech companies, demand for their services remains strong. While some areas like video conferencing are decelerating as adoption nears completion, others – such as enterprise software generally, cloud migration, and cybersecurity – are soaring (Figure 3).

Moreover, there are solid reasons why they should: they are not only increasingly business critical, but can also help corporate users improve their services and minimise costs and risks as enterprise software enhances productivity and therefore still represents an indispensable investment with a high ROI.

Figure 3
2023 tech spending in selected categories

Source: Gartner for 2023 category spending, as of 28 November 2023; company filings for 2022 revenues.

To put these numbers in some kind of context, we added the 2022 total revenues for some of the world’s largest technology companies. These categories are already rivalling trillion-dollar businesses in their earning power – and, we would note, they are generally higher margin than retailing or hardware manufacturing.

Add to this the prospect of AI surpassing them all. Even small increments in adoption across several use cases could result in phenomenal growth rates over the next few years (Figures 4 and 5).

Figure 4
Generative AI adoption rates

Source: Menlo Ventures, as of 28 November 2023.

The likely economic benefits make this increased adoption highly appealing: an academic study reported generative AI yielded a 9% improvement in the time spent handling customer issues at a company with 5,000 agents1; McKinsey suggests generative AI could offer a productivity gain worth $60-110 billion a year in pharmaceutical and medical-product industries by easing the resource-intensive process of discovering new drug compounds2.

Overall, the consultancy argued that combining generative AI with all other technologies could add 0.2 to 3.3 percentage points annually to global productivity growth, adding the equivalent of $2.6 trillion to $4.4 trillion annually to the world’s economy.

Figure 5
Generative AI spending

Source: IDC, as of 31 October 2023.

An interesting development we detected is in how AI is being priced. Whereas the principle of Software as a Service (SaaS) was to charge subscription fees on a recurring basis every month, AI-based applications – ‘intelligence as a service’ – often feature charging on a usage basis. This is because the cost of generating content (i.e. computing power) is high and can be related directly to the value and quantity of the output.

This may boost adoption by providing an easier entry point for new users: they can try it for individual projects, and then expand as new use cases emerge, rather than being dissuaded by higher upfront costs for a monthly subscription before experiencing its benefits. Indeed, many companies use a hybrid model: subscriptions give good revenue visibility, complemented by consumption-based pricing to scale as customer use increases.

What VC sees

Combining the lessons learned and tech’s undimmed potential in a portfolio is the next challenge for investors.

All the investors we heard from agreed that mistakes were made in the 2020-21 era, but they had reached different conclusions. The least hands-on actors are moving on to the next big opportunity – which is AI – and investing little energy on historical deals done while still having large funds to deploy in this new environment. Others are actively monitoring, managing and supporting portfolio companies in their transition from a ‘growth at all costs’ mentality to efficiency.

It is clear that traditional financial theory will prevail, and that companies will eventually have to post profits at some point in their journey. Those that hit the markets through less disciplined routes, like SPAC IPOs in particular, are being punished by public investors severely.

With money more expensive, and a shift from ‘need to grow’ to ‘need to survive’, companies and founders have no choice to focus on efficient growth. Fortunately, several companies – often with investors in the second camp we mention above – are succeeding in managing this transition well and demonstrate the strength of their models by maturing from burning cash into producing high profits, with EBITDA margins of 50% or more. This can and does happen very fast, but what the winners seem to have in common is strong backing from leading investors.

With this improved profitability, many managers we met are now focusing on generating distributions from their existing portfolios; so far they have preferred venture buyouts and secondary sales while the IPO window has been broadly closed (notwithstanding some high-profile listings like ARM, Instacart, Reddit, and most recently Rubrik).

Managers know they will have to demonstrate a record of returning cash to investors if they want to raise new vintages.
— Stanislas Chanavat, Principal Thematics - Private Equity, Pictet Alternative Advisors

The managers know they will have to demonstrate a record of returning cash to their investors if they want to raise new vintages, so some groups are even putting in place teams to focus on the monetisation of portfolio assets as well as ‘reinvestment’ teams to make sure that additional equity going into existing deals is assessed in an independent manner. This helps mitigate the risk that some deal partners who fell in love with their companies may want to support them at all costs, even when it would not be rational to put in additional equity.

Across the industry, we have seen some flat rounds or convertible structures put in place to extend companies’ lifetimes, but most of the venture/growth players in our own universe have been extremely selective and disciplined when deciding where to allocate additional equity in this environment.

This context has further illustrated the importance of building diversified exposure to different maturity profiles: early-stage venture capital, venture buyouts and ‘take privates’ have been most attractive in the recent past, while pre-IPO and late growth styles have generally struggled most.

What’s next?

Taken together, all of this suggests to us that we have landed in a healthier environment, where technology is still exhibiting resilience by helping the economy become more productive and the best managers have adapted skilfully. As for the period ahead, we brought home several messages from our meetings across the US, Europe and Asia:

  • Software spending continues on a strong trajectory and successful companies continue to grow at scale, but valuation levels have come down by as much as 40-50% from their 2021 peaks.
  • Deal volume has normalised, with venture dry powder still very high; we expect this to be invested at much better valuations, with less competition from ‘tourists’ that have returned to their core strategies.
  • The ability of managers to adapt to new environments is crucial; a strategy that privileges investments with hands-on investors, who can lead rounds and influence companies’ governance, is paramount in more difficult market conditions.
  • AI is at an initial stage of what is likely to become an enabling technology across all sectors of the economy, but investors will have to be cautious when it comes to deploying capital in what is likely to be an overhyped initial wave; we prefer to focus on solid technology and defensible business models.


In major platform shifts, the ultimate winners are not crowned in the early years. Think of Nokia and BlackBerry in mobile phones, Yahoo and Lycos in search, or Myspace and Friendster in social media. This again makes the case for diversification.

In turn, diversification is aided by the switch to a buyers’ market after the exuberance of 2020-21. The best managers can invest in solid business fundamentals, at great valuation entry points, having performed extensive due diligence and negotiated favourable terms.

“The Nonprofit’s principal beneficiary is humanity, not OpenAI investors.” So OpenAI insists of its non-profit legal structure. Perhaps. Today’s investors in AI and technology more generally nevertheless have good reason to expect to benefit. The industry’s return to profit discipline, complemented by the longstanding secular tailwinds for best-in-class technology, give us great confidence.

[1] “Generative AI at Work”, Brynjolfsson, Erik and Li, Danielle and Raymond, Lindsey R, National Bureau of Economic Research, Working Paper 31161, April 2023
[2] “The economic potential of generative AI: The next productivity frontier”, McKinsey, June 2023
Please confirm your profile
Please confirm your profile to continue
Or select a different profile
Confirm your selection
By clicking on “Continue”, you acknowledge that you will be redirected to the local website you selected for services available in your region. Please consult the legal notice for detailed local legal requirements applicable to your country. Or you may pursue your current visit by clicking on the “Cancel” button.

Welcome to Pictet

Looks like you are here: {{CountryName}}. Would you like to change your location?