Investing in Generative Artificial Intelligence & the Gartner Hype Cycle: Is the Hype Worth It?

“We simply attempt to be fearful when others are greedy and to be greedy only when others are fearful.” Warren Buffett, 1998 Berkshire Hathaway annual report

Is now a good time to invest in generative artificial intelligence?

Generative AI is currently the hottest province in the technology realm. ChatGPT3 reached over 100 million users within two months of its release in November 2022, faster than any application in history. Generative AI companies raised $21 billion in 2023, about 7.5% of global venture funding and four times more than in 2022. Generative AI is still a nascent category yet 36 companies have already reached unicorn status. Nvidia, a maker of semiconductors that power AI applications, is the third most highly valued firm globally behind Apple and Microsoft after its stock rose eight times since the launch of ChatGPT3 to $2.3 trillion.

Warren Buffett advises investors to be fearful when others are greedy and greedy when others are fearful. Yet the Gartner Hype Cycle suggests investors do the opposite. As I described in Seeing Clearly: Three Startup Strategies to Survive & Thrive the Gartner Hype Cycle, the venture industry is subject to flights of fancy and fits of depression as investors clamor into the fray during the Peak of Inflated Expectations and disappear in the Trough of Disillusionment.

What explains this schizophrenic behavior? The Power Law observes that a few winners typically drive 90% of early-stage returns in top performing funds. Thus, markets condition early-stage investors to compete for the most promising deals at most any price and withhold funding when performance lags or the luster fades.

The software industry offers a striking example of this trend. Investors clamored into software startups in the late 1960s following a spate of early initial public offerings[i]. But software funding evaporated in the early 1970s when early entrants faltered, and the software industry seemingly stalled for a decade until the advent of personal computers in the early 1980s. Yet the venture industry largely missed the ultimate software winners as Microsoft, Oracle, SAP and SAS were founded in 1972 to 1977 during the downturn.[ii]

Based on my review of Gartner Hype Cycle annual reports since 1995, most of the 300+ emerging technologies they tracked have followed a pattern consistent with the software industry. About 70% of emerging technologies that reached the Peak of Inflated Expectations dropped into the Trough of Disillusionment. Only 30% avoided a trough before achieving widespread adoption. Success seems inevitable as funding is plentiful and expectations are high at the peak. But expectations often outpace reality. When reality and expectations collide, interest evaporates, markets consolidation, and success seems unlikely.

However unlikely success may seem at the time, Troughs of Disillusionment delay but do not diminish prospects for emerging technologies. As Figure 1 illustrates, technologies in the Trough of Disillusionment have a higher success rate than those at the Peak. Troughs give companies time to refine products and adjust business models to achieve Product Market Fit. As troughs winnow competition, surviving companies have a higher likelihood of success than those vying in a crowded field at the Peak. Examples of emerging technologies that successfully traversed the Trough of Disillusionment include blogging, cloud computing, desktop videoconferencing, digital cash, mobile devices (PDAs), online video, SaaS software and tablets.[iii]

Figure 3: Gartner Hype Cycle Emerging Technologies — Success Rate by Stage[iv]

Investing Across the Gartner Hype Cycle: Is the Hype Worth It?

The Gartner Hype Cycle reinforces Buffett’s admonition about fear and greed. Investors with the patience and fortitude to invest in the Trough of Disillusionment instead of at the peak typically enter at better prices in more established businesses vying in markets with fewer competitors. Surviving companies are typically less capital intensive having cut costs and improved unit economics to play through the downturn. Exit horizons are typically shorter for companies that have survived the downturn.

Each of these factors suggest a higher expected investment return for companies at the trough than at the peak. Returning to our software example, Microsoft, Oracle, SAP and SAS achieved far better outcomes than their most promising predecessors Computer Sciences Corporation, Boole & Babbage and Applied Data Research.

Troughs of Disillusionment are also surprisingly good times to start companies. Startups founded during downturns are not burdened with cost cutting or technical debt. They also have more access to experienced talent and more time to perfect products and business models with less competitive noise. Some of the most successful tech companies were founded during downturns. Startups founded during the financial crisis from 2009–2011, for example, include Instagram, Pinterest, Slack, Snapchat, Square, Stripe, Uber and WhatsApp. Other iconic technology companies that rose to prominence during downturns include Apple, Cisco and Google.

Yet venture investing remains highly cyclical. Global venture outlays in 2023 were at the lowest level since 2017 and about 40% the level in 2021. And yet 2021 was a good time to have invested as exits were at an all-time high, while 2023 was more likely a good time to invest.

If prudent venture investing is countercyclical, why does it remain highly cyclical? There are many potential explanations, yet three factors weigh heavily from a fund manager’s perspective:

· Funds have access to more capital during bull markets and less during downturns as limited partners who fund venture firms make commitments based on liquid capital, which is highly correlated with exits.

· Established funds manage large portfolios, which require increased attention during downturns. Thus, funds tend to shift focus from new investments to portfolio triage during downturns.

· As we know from behavioral economics, investment psychology is procyclical: investors wax confident in bull markets but are loss averse during downturns.

Oases appear in the desertic Trough of Disillusionment offering opportunities for both established investors and emerging managers. Downturns reward patient capital, which may enter at more reasonable valuations while selecting from among the most promising startups that have survived the downturn. Downturns also favor emerging managers, which may invest in the next generation of startups while established investors tend to their portfolios.

Islands of rationality also emerge in the torrential Peak of Inflated Expectations. Bull markets generally raise valuations, yet gaps often emerge in overlooked sectors as investors flock to the hottest markets.

As a result, top venture firms increasingly take a barbell approach with early stage funds to invest at the Technology Trigger and growth stage funds to invest at later stages.

So, we return to our initial question: Is now a good time to invest in generative AI?

Gartner was prescient as generative AI first appeared on the Hype Cycle in 2019. The Gartner Hype Cycle proved a harbinger of opportunity as generative AI ascended the Technology Trigger stage in 2020 and 2021, approached the peak in July 2022[v], and is perched atop the Peak of Inflated Expectations since OpenAI’s release of ChatGPT3 in November 2022. The venture industry noticed investing nearly $10 billion in generative AI businesses in 2021.

Generative AI is a transformational technology with broad applications that will emerge over an extended time horizon. Investing wisely now at the Peak of Inflated Expectations is tricky and requires unique insight in niche market segments as foundational models have significant scale requirements that favor incumbents and highly capitalized players. An alternative approach is to invest in overlooked segments that may benefit from generative AI as the technology expands and matures. Otherwise, there will be future opportunities to invest when expectations have moderated.

[i] Early public software companies founded in the 1950s and 1960s included Applied Data Research, Advanced Computer Techniques, Boole & Babbage, Computer Sciences Corporation, Cullinet, Informatics General, International Data Group, and System Development Corporation.

[ii] Software history source: Martin Campbell-Kelly, From Airline Reservations to Sonic the Hedgehog: A History of the Software Industry, MIT Press, 2003. I applied the Gartner Hype Cycle, which started in 1995, to the software industry in the 1970s for this example.

[iii] Technologies in the Trough of Disillusionment: in some cases, I have updated the technologies that have rebranded to reflect current usage rather than names used at time.

[iv] Gartner does not publicly report on the ultimate success of emerging technologies that appear on the Gartner Hype Cycle. I kept a scorecard for all emerging technologies listed by Gartner from 1995 to 2015 anticipating that nine years would be sufficient time to ascertain a degree of success. For listings prior to 2015 that Gartner no longer tracks in the Hype Cycle, I placed technologies in one of three categories: (1) Startup Success indicating that at least one startup achieved significant success based on the technology; (2) Incumbent Success indicating that established firms used the technology to significantly expand an existing business or expand into a new segment; and (3) Not Impactful in the absence of #1 or #2. Success rates are based on (1+2)/(1+2+3), which are reported in Figure 5.

[v] The Gartner Hype Cycle referred to Generative AI in 2019 to 2021 and Foundational Models in 2022.