The Coming Shakeout in Supply Chain Tech: $80B at Risk in the Age of AI.
By Dweep Chanana and Wolfgang Lehmacher
This perspective combines Dweep Chanana’s experience in strategic investing and Wolfgang’s deep expertise in supply chain technology.
Since 2018, we have seen countless startups emerge to solve the same enduring problems in supply chain management and logistics — pricing inefficiencies, lack of visibility, delayed and inaccurate data, reliance on manual data entry, human error, and poor responsiveness to disruptions. Billions have flowed into the sector, driven by the promise that solving these problems step-by-step would create durable value.
The rapid evolution of generative AI, large language models (LLMs), and agentic AI is fundamentally challenging the assumptions under which that money was invested and risks stranding billions in invested capital. This has stark implications both for founders and shareholders that had anticipated near-term exits and for investors looking today to enter the space.
This article draws on insights from our own experience and investment, especially in Roambee, as well recent discussions with other investors and corporates on how to reposition businesses for an AI-first startup world and what truly constitutes differentiation.
The old growth model: from visibility towards automation
Startups in the sector have often followed a familiar staircase. They typically began by giving customers a clearer picture of what was happening in their supply chain (visibility). With funding, they expanded into prediction (forecasting demand, ETA prediction, or disruption risk) and then decision support, offering recommendations to optimize operations.
The eventual goal has been autonomy — not unlike self-driving cars, where the leap is from driver-assist to fully autonomous navigation. Companies like Project44, FourKites, and Shippeo (visibility providers) and Forto, Flexport, and Freighthub (freight aggregators) all followed this trajectory, raising hundreds of millions of dollars each. Most recently, project44 released an Intelligent TMS, to move up the stack and offer transportation execution and inventory management.
Each step up this stack promised higher margins, stronger customer lock-in, and a more compelling story for acquirers. The assumption was that the time and technical difficulty required to climb the staircase would translate into durable defensibility. But that logic no longer holds.
How AI changes the software landscape
At their core, most enterprise and supply chain software solutions do two things: ingest data and apply rules or algorithms to analyze it. AI tools are making both steps dramatically easier, collapsing the time and cost required to build capabilities that once took years. What startups thought of as their moat — the hard-won progression up the staircase — is now becoming a commodity.
For example, in supply chain visibility, FourKites, Project44, and Shippeo together raised $1.3bn to build large, aggregated visibility datasets. New entrants can now stand up similar capabilities in months using AI-enabled tools. The result is rapid commoditization.
The same is happening on the analysis and prediction side. AI-driven optimization libraries allow companies to build optimal freight matching between shippers and carriers – a fundamental capability of digital forwarders that collected attracted over $20bn in funding. Demand forecasting is also not immune: off-the-shelf AI models from Amazon (Forecast), Azure, and open-source Prophet, allow integration of historical and inventory data to receive forecasts with minimal customization. This once took specialist teams months or years to integrate and refine.
AI also reduces acquirer and customer appetite for 3rd party solutions, as acquirers can now theoretically build in months what once took years. McKinsey estimates generative AI can boost developer productivity by 35–45%, halve documentation time, and cut code refactoring by up to 30%. In our view, startups’ biggest competition in the industry has always been “do-it-yourself” views, usually held by a corporate CIO, and shorter development times strengthen this “build vs. buy” argument.
The $80 billion at risk
For investors, the exposure is significant.
According to McKinsey, more than $81.3 billion was invested into logistics and supply chain technology startups globally from 2018 to 2023. Those same reports suggest that about half of the invested capital ($40bn) was into asset-light categories (visibility platforms, digital aggregators, TMS/orchestration, software-enabled services) whose value proposition is most at risk of erosion from AI. Today, much of that capital is at risk of being stranded or written off.
The problem is compounded by a sharply narrowing financing & exit market.
Equity investments in the sector fell by 90% to just $2.9bn in 2023, even prior to the emergence of AI.
Merger and acquisition (M&A) momentum has also slowed dramatically: platform transactions fell by 47.5% from 2022–H1 2025, with just 21 deals completed, compared to 40 in the 2019–2021 period.
Finally, IPO activity has also fallen - only 10 public offerings took place since 2022, compared to 18 in 2021 alone.
Exit options are also complicated by changing views of strategic acquirers. As mentioned previously, acquirers can reasonably now ask themselves if it is better to build capabilities than to acquire. This is one reason why acquisitions within the sector have dried up. As I discussed with a shareholder in System Loco, exit strategies are usually oriented towards potential buyers’ growth plans, offering them missing capabilities. When the strategies of buyers themselves are in flux, who or what do you orient to?
In this environment, the central question is no longer how much capital has been raised by a startup, but how to defend that investment. The answer will determine whether investors achieve exits or are forced to write-off their investments.
Where is the differentiation for startups?
In this new world, technology is no longer a moat. Real defensibility comes from assets and capabilities that cannot be quickly copied — qualities that investors should use as filters when evaluating both existing holdings and new opportunities.
Proprietary data is an advantage – Exclusive, hard-to-replicate datasets that make a company’s AI outputs consistently better than competitors’ are key. Without a data edge, algorithms converge to the same baseline. This may explain why Overhaul recently bought FreightVerify, to secure access to underlying visibility data.
Operational context is the moat – Deep, tacit understanding of customer workflows, constraints, and industry quirks is arguably one of the most durable USPs a startup can have today. This was confirmed recently by Ossi Tianen, Partner at NGP Capital, on why they invested in Tractian, which is building an “industrial copilot for manufacturing AI”. We also see the opposite challenge with AI-first startups that struggle to find product-market fit, because they have built a broad solution looking for a problem to solve. Understanding the industry problem thoroughly is crucial for building AI-enabled solutions that customers are willing to pay for.
Leadership capabilities are key – In a world where the marginal cost of developing technology is low, what would you have it do? And how do you reorient a company for such fundamental change? Getting the answer right will depend on the quality of leadership in companies and their ability to foresee and oversee change.
Case in Point: Roambee’s continuing evolution
Roambee began as a visibility platform in a segment now rapidly commoditized by AI-powered ingestion and parsing. Anticipating this shift very early, Roambee pivoted toward a model anchored on deeper customer problem-solving, leveraging its sensor network and domain expertise to address operational decision-making rather than just monitoring. This move positioned it ahead of the curve — and exemplifies the kind of agility and customer intimacy investors should now prize.
Conclusion
The emergence of artificial intelligence is turning hard-earned capabilities into commodities. It also reduces acquisition appetite, as buyers adjust to a new landscape of possibilities. This risks the billions already invested in the sector, as well as new investments being made today, and challenges previously reliable exit strategies.
In this new landscape, the basis of competition shifts from technology to data, context, and integration. To adjust and benefit, startups in the sector must rethink their growth and exit pathways. Investors must quickly pivot companies and capital to match. The challenge is to root oneself deeply in customer operations and build moats that AI alone cannot breach.
The critical question is: can the existing startups — into which billions have already been invested — grow quickly and embrace AI fast enough to avoid commoditization and maintain or even increase shareholder value?
