Morgan Stanley mapped 3 AI futures. One company wins them all.

Will the AI early beryllium closed aliases open, aliases a substance of both? Morgan Stanley studied these 3 scenarios and recovered that immoderate companies will triumph nary matter what happens.

In a investigation statement this week, the bank's strategists laid retired 3 imaginable paths for the AI marketplace complete the adjacent fewer years.

The winners alteration depending connected which script plays out, but 1 taxable stays remarkably consistent: The companies building the computing backbone that runs AI models and services will stay successful a beardown position regardless.

Closed wins

The first script is 1 wherever closed, proprietary AI models support their lead. Companies proceed paying a fistful of frontier AI labs because they connection the champion performance, security, and easiness of use.

(The slope doesn't officially screen OpenAI and Anthropic, because they're still backstage companies, but I fishy these would beryllium the main winners here).

This result would besides favour awesome unreality providers including Amazon and Google, on pinch companies supplying the chips, networking equipment, and energy needed to tally monolithic AI systems, specified arsenic Nvidia and Bloom Energy.

A hybrid world

The 2nd anticipation is simply a hybrid world, which Morgan Stanley views arsenic a realistic outcome. In my experience, these much nuanced outcomes are usually what ends up happening.

In this scenario, businesses would usage premium, closed AI models for the hardest jobs while relying connected cheaper, customizable models for regular work. That would dispersed AI crossed nationalist clouds, backstage information centers, and company-owned hardware, creating opportunities for unreality providers, infrastructure software, and cybersecurity firms.

Winners would see hyperscale unreality providers specified arsenic Amazon, Google and Microsoft. Software companies including Datadog, Palantir, Crowdstrike, Okta, and ServiceNow would besides do good successful this world. Oh, and Nvidia, too.

Open wins

The 3rd script assumes unfastened models go conscionable arsenic tin and easy to usage arsenic proprietary ones. Lower costs would promote companies to deploy AI overmuch much broadly, pinch much workloads moving wrong their ain information centers aliases connected section devices alternatively of centralized unreality services.

That would use makers of endeavor hardware, information software, and alleged separator computing instrumentality which runs smaller systems person to users.

Microsoft would still do good successful this world, according to the Morgan Stanley strategists. But caller winners would beryllium Chinese AI labs and China's tech giants, specified arsenic MiniMax, Z.ai, Alibaba and Tencent.

Dell, HP, and Apple would besides thrive successful this script because they specialize successful making devices that tin tally AI models and services person to extremity users. Oh, and Nvidia wins again here.

Constant winners

Despite these very different outcomes, Morgan Stanley recovered immoderate clear constants.

You've already spotted the clearest winner: Nvidia appears among the biggest beneficiaries successful each 3 scenarios, reflecting the continued request for AI servers and networking nary matter which type of package wins.

The unreality giants besides characteristic prominently, though their rankings displacement depending connected whether proprietary aliases unfastened models summation the precocious hand. Microsoft stands retired successful the open-model scenario, while Amazon and Google are strongest if proprietary models stay ascendant aliases the marketplace settles into a hybrid approach.

Morgan Stanley's broader conclusion is that the statement complete which type of AI exemplary yet wins whitethorn matter little than galore investors think.

If AI take continues to accelerate, the companies supplying the infrastructure that powers it could prosper successful almost immoderate future.

Sign up for BI's Tech Memo newsletter here. Reach retired to maine via email astatine [email protected].