Winter may have come early for AI and machine learning startups. After years of rosy forecasts, growth and investor enthusiasm, a fresh one has emerged report from PitchBook shows that VC activity in the AI sector has declined dramatically over the past few months.
Deal value growth in AI startups declined 27.8% quarter-over-quarter in Q2 2022, with total investments reaching $20.2 billion across 1,340 deals. Year-to-date, VCs have given $48.2 billion to AI startups across more than 3,000 deals — which sounds hearty, but actually represents a 20.9% year-over-year decline.
Financing artificial intelligence rejected at all stages of the transaction, according to PitchBook data. Excluding angel and seed rounds, early-stage investments totaled $4.2 billion, down from $5.6 billion in the first quarter and down 35% from the same quarter last year. Meanwhile, later-stage funding dropped from $18.3 billion in the first quarter to $13.4 billion in the second quarter, a 48% decline from a year ago.
As the value and number of VC deals in AI startups hit their lowest levels since the fourth quarter of 2020, it’s not just investors who are retreating.
According to a recent one questionnaire The Harris Poll and Appen found that fewer companies are committing $500,000 to $5 million budgets to AI, choosing instead to augment capital commitments to overall cloud computing and edge infrastructure.
“In the second quarter, investors all but halted funding in leading horizontal categories, including artificial intelligence platforms and semiconductors,” PitchBook senior analyst Brendan Burke, author of the PitchBook report, told TechCrunch in an email interview. “With AI in particular, enterprises have reduced their appetite for specialized AI infrastructure, opting to reuse existing infrastructure to support AI workloads for models that have already proven to be effective. This pressure limits startups’ ability to transform their customers’ data stacks. “Moreover, specialized chips are not seeing much success in this environment after successful pilots in 2021, and enterprises are content to leverage existing compute clusters.”
Burke didn’t immediately write off every category of AI startups, pointing out that some operate cases – particularly accounting automation and wealth management – have proven resilient in the face of the current economic downturn. At the same time, he noted that many AI platform companies are struggling to meet their revenue projections and are competing with incumbents such as Google, Amazon and Microsoft for a relatively diminutive market to reach. Burke estimated the market size for AI software platforms at $7 billion – a modest amount compared to, for example, $17 billion HR software market
“The 2021 IPO wave came and went without significant impacts for horizontal AI startups, leaving questions about the size of the AI software market and the opportunities for AI chipmakers,” Burke said. “Chip companies offer exciting innovation, yet a high-risk scale path that is difficult to provide to mainstream investors.”
When it comes to top-tier AI funding numbers, there’s more bad news. A PitchBook report shows that investment in artificial intelligence as a service will decline by 87.7% in 2022. In terms of deal volume, VC exits in AI declined 21.8% as M&A activity remained subdued, apart from IBM’s acquisition of database management startup Databand and Meta’s purchase of recommendation company Presize.AI clothing sizes.
However, Burke said there is reason for optimism regarding specific segments of AI, such as data preparation, computer vision, robotic process automation, natural language processing and substantial data analytics technology.
“In general, investors are currently biased towards the capital plans of resource-intensive AI startups due to fiscal discipline, but we expect their confidence to be restored given the successes in this space. The widespread shift to AI-specific infrastructure remains a long-term trend,” he continued. “Database management [also] continues to grow rapidly as enterprises move to non-relational databases and data lakes. Spending in these areas is now expected to exceed relational database management by 2026, which we believe will be a boon for AI decision engines given their ability to analyze unstructured data from a variety of sources.”
What does this mean for AI startups? Burke recommended that they strive to leverage existing infrastructure and integrate with systems of record to find their place in the data analytics stack. Overall, he said, vendors building analytics tools on top of leading cloud database systems and industry-specific logging systems can expect faster implementation.
“The goal of an end-to-end platform requires large investment and has limited product market fit,” Burke added. “Vertical applications such as sales and marketing, drug discovery and information security rely on existing regulatory systems and continue to drive strong growth. To restore confidence, horizontal platforms will need to demonstrate exit paths in 2023.”