What began as one of the most closely watched acquisitions in the artificial-intelligence industry has ended in a dramatic reversal.
Anthropic, the company behind Claude, has abandoned plans to acquire Israeli AI startup Decart after conducting due diligence on the three-year-old company. The proposed transaction had reportedly been valued at roughly $6 billion, and at one point appeared close to completion.
The collapse is significant not simply because a multibillion-dollar acquisition has failed, but because of what Decart represents in the next phase of the AI race: making artificial intelligence cheaper, faster and more efficient to operate.
From Nvidia to Anthropic—and back to uncertainty
Decart's rise has been extraordinarily fast.
Founded in 2023 by Dean Leitersdorf and Moshe Shalev, the company employs roughly 100 people and has raised about $450 million. Its most recent financing valued it at approximately $4 billion.
The company attracted serious attention from Nvidia, which was not only an investor but also reportedly held advanced acquisition discussions with Decart.
Then Anthropic entered the picture.
Decart halted its Nvidia negotiations after Anthropic made its offer, with the proposed Anthropic transaction eventually reported at between $6 billion and $7 billion.
For a startup only three years old, this was extraordinary.
But after Anthropic conducted deeper due diligence, the deal collapsed.
What did Anthropic discover?
This is the most important unanswered question.
There is currently no public evidence establishing exactly what caused Anthropic to walk away. CTech notes that the problem could potentially involve Decart's technology, scalability or business strategy, but these remain possibilities rather than confirmed explanations.
That distinction matters.
Decart's technology is designed to improve AI computing efficiency, potentially allowing models to run faster and reducing the enormous cost of inference.
On paper, that makes Decart strategically attractive to Anthropic.
Claude's growing demand requires enormous computing resources. Every improvement in computational efficiency can potentially translate into lower operating costs and greater capacity.
But a technology that works impressively in demonstrations or limited deployments is not necessarily a technology that can deliver the same gains across billions of dollars of computing infrastructure.
Due diligence exists to answer precisely that question.
Anthropic may have concluded that the technology did not justify a $6 billion acquisition—or that it could achieve similar results internally or through a partnership.
We simply do not know yet.
The biggest lesson: AI's next war is about economics
The Decart story reveals something important about where the AI industry is heading.
The first phase of the AI race was largely about who could build the most capable models.
Now the competition is increasingly about who can operate those models most efficiently.
Training frontier AI systems costs enormous sums. Serving millions of users creates another enormous expense.
This means AI companies are increasingly interested in optimization, inference efficiency, specialised chips, data-centre architecture and software that can extract more performance from existing hardware.
That makes companies like Decart strategically important.
The question is no longer simply:
"Who has the smartest AI?"
It is increasingly:
"Who can deliver intelligence at the lowest cost?"
Why Nvidia's interest makes sense
Nvidia's interest in Decart was particularly logical.
Nvidia dominates the AI accelerator market, and software that can make computing hardware perform more efficiently has enormous strategic value.
Decart has reportedly developed technology designed to work across different chip architectures rather than being limited exclusively to Nvidia GPUs. Its capabilities include AI infrastructure optimisation as well as real-time video generation.
For Nvidia, acquiring such technology could strengthen its position beyond simply selling chips.
It could help Nvidia sell an entire optimisation ecosystem.
That is why the company's reported willingness to acquire Decart is significant.
For Decart, the collapse changes everything
The failed Anthropic deal leaves Decart in an unusual position.
Just months ago, investors valued the company at approximately $4 billion. It then appeared to have a potential buyer willing to pay substantially more.
Now that transaction has disappeared.
That does not mean Decart is a failed company.
Far from it.
It remains backed by major investors and possesses technology that attracted some of the world's most powerful technology companies.
But the failed acquisition inevitably raises questions about its independent business model.
CTech reports that Decart has generated revenue in the tens of millions of dollars, while its business model has continued evolving.
That gap between technological promise and sustainable commercial scale is one of the defining problems of today's AI startup ecosystem.
A company can be worth billions on the basis of talent, technology and strategic potential long before it has demonstrated a durable business capable of supporting that valuation.
Israel's AI ambitions are also at stake
There is a broader national dimension.
Decart became one of the most prominent symbols of Israel's ambitions to become a major force in artificial intelligence.
An Anthropic acquisition would have brought one of America's leading AI companies directly into Israel's technology ecosystem and potentially created a major development centre there.
The collapse does not erase Israel's AI capabilities.
But it does demonstrate how difficult it is to transform a strong concentration of engineering talent into globally dominant AI companies.
Israel has enormous technological expertise, particularly in cybersecurity, defence technology and advanced computing.
The challenge is turning that expertise into companies capable of building massive, sustainable global businesses rather than becoming acquisition targets for American technology giants.
Collaboration may still happen
Interestingly, Anthropic's decision does not necessarily mean it has concluded that Decart's technology is worthless.
The companies could still collaborate, according to reports.
That may ultimately prove more rational for Anthropic.
Instead of spending $6 billion to own Decart, Anthropic could potentially license its technology, enter a strategic partnership or work with the startup on specific optimisation problems.
For Anthropic, that preserves flexibility.
For Decart, it preserves independence.
And for investors, it keeps open the possibility of another acquisition in the future.
The real winner may be whoever solves AI's cost problem
The Decart saga ultimately exposes the next battlefield in artificial intelligence.
The industry has already demonstrated that increasingly powerful models can be built.
Now comes the harder question:
Can they be made cheap enough to become infrastructure for the entire global economy?
If AI inference remains extraordinarily expensive, only the richest technology companies will be able to operate frontier systems at massive scale.
If companies can dramatically reduce those costs, AI becomes accessible to millions of businesses.
That makes AI optimisation potentially as important as AI model development itself.
Anthropic's decision to walk away from Decart therefore should not be interpreted simply as a rejection of an Israeli startup.
It may instead be a sign of a maturing industry becoming much more disciplined about what deserves a multibillion-dollar valuation.
The $6 billion deal may have collapsed.
But the race to make AI faster, cheaper and more efficient has only just begun.
Comments
Post a Comment