-
There are multiple factors which influence the decision to build vs buy-in to AI technology, analysts told Silverlinings
-
Partnerships are likely to continue, but expect hyperscalers to maintain an arms length from collaborators, they said
-
The analysts also tipped consolidation of AI startups to occur over the next three years or so
Microsoft and OpenAI. Amazon and Anthropic. But also Bard, Bedrock, Duet and Vertex. As the artificial intelligence (AI) landscape rapidly evolves, it seems cloud hyperscalers are taking a hybrid approach to the technology, both developing their own solutions and partnering with leaders in the space. But how, exactly, are they deciding what to build vs buy?
We put the question to a series of analysts. Here’s what they had to say.
“Cost, AI maturity, [time to market] TTM and core business focus will each factor into the decision to ‘buy’/partner or ‘build’ generative AI capabilities internally,” Reece Hayden, senior analyst at global technology intelligence firm ABI Research, told Silverlinings.
“Generally, ‘buy’ strategies offer high Capex, but lower Opex, higher TTM for less mature players. They align more effectively with those trying to drive cloud infrastructure usage or those integrating AI capabilities into core products/services,” he continued.
“On the other hand, ‘build’ strategies have lower upfront costs, but higher OpEX and will be pursued by mature AI companies with a core focus on managed services, consultancy, system integration; rather than cloud infrastructure usage.”
Hayden said that all cloud players will partner with AI specialists to some extent, with Nvidia a key target for its hardware capabilities. But too much partnering could backfire, he added. That’s because it could hamper differentiation and thus hinder competition.
Dell’Oro Group VP Sameh Boujelbene offered a similar take, citing the time-to-market and cost considerations that go into the build vs buy equation. “As a response, we anticipate that Cloud Service Providers will adopt a hybrid approach, balancing in-house development with external acquisitions, as they strive to meet their targets.”
Meanwhile, Sid Nag, VP of Gartner’s Technology and Service Provider group, concurred hyperscalers will likely limit partnership in certain areas to avoid exposing their “crown jewels” to third party companies. By way of example, he pointed to Oracle’s recently announced collaboration with AI company Cohere.
Oracle said it will integrate Cohere’s technology into its Oracle Fusion Cloud Applications, Oracle NetSuite and other industry-specific applications “but they never actually talked about how they’re going to use it for Oracle Autonomous Database,” Nag noted. Since Oracle Autonomous Database is one of the company’s “crown jewels,” Nag said they will likely use their own AI technology for that and keep Cohere at an arm’s length.
What’s next
Looking ahead, Hayden said consolidation is likely to occur over the next three years or so, as “capital-rich players” scoop up generative AI startups.
He added that ABI Research expects the enterprise market to align behind “open, tailored, fine-tuned models” with between 1 billion and 15 billion parameters which allow for customization.
“This means that AWS’ approach is likely more robust in the long run compared to GCP, IBM or Microsoft,” he said.
Nag said that in the short term, AWS has some work to do in explaining how its AI tools can be applied by enterprises. He noted that unlike Microsoft and Google Cloud, who have demonstrated how AI can be incorporated into their respective suites of enterprise applications, AWS doesn’t have a clear application strategy – yet.
“I never discount AWS,” Nag concluded. “They always have something up their sleeve.