Bloomberg Tech

Will Manidis on Why AI's Winners Own the Rails

Will Manidis· SVP of Veradigm AI at Veradigm
·~6 min·English·Bloomberg
Business StrategyAI CompanyInference
TL;DR

As the cost of frontier intelligence collapses, Manidis argues the lasting economic gains flow not to the labs that build AI but to the firms that own the rails and distribute it into the real economy.

01Core Thesis

The Commanding Heights

As intelligence scales, Manidis argues <strong>the firms that control distribution — the networks and rails commerce flows across — can benefit disproportionately.</strong>

the companies that control the rails, the companies that control the distribution to take the intelligence that's developed inside of firms like OpenAI and Anthropic and diffuse it into the real world will benefit disproportionately.

— Will Manidis, Bloomberg Tech
Key Insight
If the commanding heights are distribution, then the current market story — pricing OpenAI, Anthropic, and NVIDIA as the prize — is watching the layer least likely to keep the gains. The scarce asset stops being the model and becomes the channel into the economy.
02The Premise

Intelligence Is Getting Cheap, Fast

<strong>Frontier intelligence costs about 13x less than it did twelve months ago</strong>, the cheap-and-abundant premise the whole thesis rests on.

the cost of an individual unit of Frontier Intelligence, that is a token that is at the Frontier being used to power workflow, has fallen about 13 times over the last twelve months.

— Will Manidis, Bloomberg Tech
Key Insight
A 13x annual price drop means a capability that is a premium product today is a commodity input within a year. That inverts the usual moat logic: owning the model is owning a fast-depreciating asset, while owning the demand it unlocks compounds.
03New Category

The Neodollar

Because <strong>intelligence is dollar-denominated, Manidis sees a new, unpriced category of commerce businesses — the neodollar — standing to benefit.</strong>

intelligence is dollar denominated. Right? AI is an American business. It's the American economy.

— Will Manidis, Bloomberg Tech
Key Insight
Calling it the neodollar reframes AI spend as monetary flow, not software spend: if intelligence is dollar-denominated, the firms that carry that commerce sit closer to the money than the ones producing the intelligence — the position a toll road holds over the cars that drive it.
04Historical Lesson

Invented Here, Captured There

<strong>A technology's inventor rarely captures its economic benefit</strong> — the printing press and the railroad fortunes paid off for others, and Manidis expects AI to be no different.

Very rarely is the place where technology invented also the place where that economic benefit is captured.

— Will Manidis, Bloomberg Tech
Key Insight
His examples are survivorship warnings aimed at today's AI investors: the obvious winner — the inventor — is usually not the economic winner. The implication is that buying the lab is the printing-press-maker's trade, not the railroad financier's.
05Market Implication

All Equities Will Reprice

Manidis argues <strong>all equities will reprice because the structure of the economy itself has changed</strong>, not just the handful of stocks the market trades bottleneck to bottleneck.

we have not adequately understood that the fundamental structure that underlies The US economy has changed. All equities will reprice as a result of this

— Will Manidis, Bloomberg Tech
Key Insight
Bottleneck trading describes a market reacting component by component because it has no model for the whole. His sharper claim locates the mispricing in the economy's structure rather than any single stock, which is why he frames the correction as all equities repricing together instead of one more trade in the sequence.
06The Archetype

Meta, the Archetype

<strong>Meta is his archetype: a firm that can both train frontier models and owns the rails to distribute and monetize them</strong>, holding the whole chain.

Meta is an incredibly interesting business because not only can it do frontier training itself, but it also owns the rails to handle the diffusion of that intelligence to the global economy and also handle commerce on the back end of it.

— Will Manidis, Bloomberg Tech
Key Insight
Meta matters to the thesis less as a stock pick than as an existence proof: a single firm can occupy the commanding heights by owning training, distribution, and commerce at once. It shows the argument isn't anti-lab — it favors whoever controls the most of the chain.