Generative AI’s supply chain money tree cannot last forever and sooner rather than later they must look to build strong and differentiated revenue strategies to offset their ever-growing operational cost burden
Generative Artificial Intelligence (AI) is not a new technology, but its democratization was triggered by the emergence of powerful public APIs, notably ChatGPT. Since then, the consumer market has experienced tremendous traffic with high usage. In the wake of this demand, the supply chain has exploded with significant activity across each layer (R&D, hardware, foundation model, ML service tools, data services, applications, and ethics/regulation), trying to build access to commercial opportunities in the enterprise segment. According to a new report from global technology intelligence firm ABI Research, this market will rapidly expand by 2030 at a CAGR of 162%, offering nearly US$60 billion in revenue for supply chain stakeholders.
“Capturing the enterprise commercial opportunity is essential as stakeholders continue to struggle beneath a significant cost crisis driven by the consumer segment. Building data sets, deploying infrastructure, training, and running large language models like Claude, LLaMa, Titan, or GPT-3.5 has a sizeable cost burden that will be challenging to reduce,” explains Reece Hayden, Senior Analyst at ABI Research. “Popular public applications like ChatGPT cost at least US$500,000 per day to operate. That cost will only rise as usage increases. So far, stakeholders have relied on external funding to support free access through venture capital investment or internal subsidies. This cannot continue forever, and stakeholders must identify strategies to start generating revenue in the consumer and enterprise segments.”
Given their high customer acquisition cost, stakeholders are primarily stuck in ‘freemium revenue models that are largely unsustainable in the consumer segment. “These models are also mostly unfit for purpose in the enterprise market,” Hayden points out. “Although moving on from this model will be challenging, the good thing is that plenty of monetization opportunities have applicability across the supply chain.” Stakeholders can look to implement advertising models like those used by search engines, revenue share models which have proven successful in adjacent areas like cloud marketplaces or even look to productize open-sourced LLMs with closed-source enterprise functions. But it is vital that stakeholders carefully align their capabilities with a revenue model as some may not be fit for purpose.
The most successful revenue generation strategies over the foreseeable future will look to support enterprise adoption directly. Most enterprises lack Machine Learning (ML) skills/tools, operational expertise, and strategic legal/governance frameworks to support generative AI development and implementation effectively. For this reason, Hayden recommends, “Supply chain stakeholders should look to provide consultancy services or build low/no-code platforms that support development, deployment, fine-tuning, optimization, operational change management, and day two operations.” Business consultants like Bain, McKinsey, and BCG have joined system integrator incumbents through partnerships with foundation model owners like OpenAI. Moving forward, the enterprise service part of the supply chain could be worth more than US$15 billion by 2030.
Other factors are at play in this market, most notably the increasing focus on data privacy, which will trigger increasing interest in data service providers. Enforcement of copyright regulation for training data and enterprise demand for fine-tuning will create sustained interest in companies able to curate enterprise datasets or generate synthetic databases. “Recent fundraising rounds indicate more significant interest in ML data companies. For example, Mostly AI, a synthetic data generator, has just raised US$25 million, while Snorkel AI recently raised US$85 million at a valuation of US$1 billion, Hayden says.
“The supply chain has plenty of opportunities to offset their cost burden with revenue models, and some are already looking to do so,” concludes Hayden. “Beyond simply identifying new revenue models, stakeholders should look to build strong partnerships across the supply chain, build products/services that target B2B deployment and scale, and develop a leading position in responsible AI development.”
These findings are from ABI Research’s Generative AI: Identifying Technology Supply Side Revenue Opportunities application analysis report.
