Berkshire vs SoftBank: A $400B Clash of AI Philosophies

Two Titans, Two Opposite Bets on the Future of AI

On May 2, 2026, the Berkshire Hathaway annual shareholder meeting in Omaha told a story not through what was said, but through what was missing. Only half of the arena's 18,000 seats were filled — a stark contrast to years past when tickets were scalped and hotels were booked months in advance. Meanwhile, 6,000 miles away in Tokyo, SoftBank Group was packaging its unprofitable AI assets into a new entity called Roze AI, targeting a $100 billion valuation and a US IPO in the second half of 2026.

These two events, happening in the same week, crystallize the most consequential investment debate of our time: in the age of artificial intelligence, is it wiser to hoard cash and wait, or to go all-in on the technology that may reshape every industry on earth?

Key Takeaways

• Berkshire Hathaway holds approximately $397.4 billion in cash, its largest reserve ever, with no clear deployment plan under new CEO Greg Abel
• SoftBank committed $64.6 billion to OpenAI, a figure that could eventually approach $100 billion
• SoftBank is creating Roze AI, a vehicle to package and list its AI assets at a $100 billion valuation
• Warren Buffett, now 95, has stepped back from actively managing the annual meeting
• The two companies represent diametrically opposed philosophies on how to navigate the AI revolution
• Only one approach is likely to be vindicated by the end of the decade

Berkshire's $400B War Chest: Patience or Paralysis?

Berkshire Hathaway's cash pile has become the elephant in every room where value investing is discussed. At nearly $400 billion, it represents one of the largest concentrations of liquid capital in corporate history — exceeding the GDP of countries like Denmark or Thailand. Investors at the Omaha meeting pressed Abel on why the company refuses to deploy this capital into the market.

The answer, rooted in Buffett's decades-old philosophy, is deceptively simple: there is nothing worth buying at current prices. Berkshire's approach has always been to wait for 'fat pitches' — opportunities where the margin of safety is so wide that the risk of permanent capital loss is minimal.

But critics argue this philosophy may have a fatal blind spot. Traditional value investing depends on the ability to assess intrinsic value based on historical earnings, tangible assets, and predictable cash flows. AI companies — with their exponential growth curves, winner-take-all dynamics, and network effects — defy these frameworks entirely. By the time an AI company looks 'cheap' by Buffett's metrics, the investment opportunity may have already passed.

The half-empty arena in Omaha may be the most telling indicator. The faithful are beginning to wonder whether Berkshire's legendary patience has become a liability in a world that moves at the speed of GPU clusters.

SoftBank's All-In AI Gambit: Vision or Madness?

Son Masayoshi has never been accused of excessive caution. The SoftBank founder, whose $20 million investment in Alibaba became worth over $100 billion, has staked his company's future — and arguably his legacy — on the belief that artificial general intelligence is not just coming, but coming soon.

The numbers are staggering. SoftBank's commitment to OpenAI alone stands at $64.6 billion, making it the single largest corporate bet on any AI company in history. To fund this and other AI investments, SoftBank is now engineering financial structures that would make Wall Street's most creative bankers blush.

Roze AI, the newly announced entity, is designed to:

• Consolidate SoftBank's various AI holdings into a single investable vehicle
• Achieve a target valuation of $100 billion before generating meaningful revenue
• List on US exchanges to tap into American investor appetite for AI exposure
• Generate the capital needed to continue funding SoftBank's OpenAI obligations
• Create a publicly traded 'AI pure play' that can issue its own equity and debt

This is financial engineering in service of a technological vision. Son is essentially building a perpetual capital machine designed to feed the insatiable appetite of AI development. Whether this is genius or recklessness depends entirely on whether AI delivers on its most ambitious promises.

The Philosophy Gap: Why Both Can't Be Right

At their core, Berkshire and SoftBank represent two incompatible theories about how the future unfolds. Berkshire's worldview assumes mean reversion — that markets overshoot, bubbles pop, and patient capital eventually finds bargains in the wreckage. This has worked spectacularly for 60 years.

SoftBank's worldview assumes technological singularity — that AI will create such an enormous step-change in economic value that traditional valuation metrics become irrelevant. In this framework, any price paid today for a leading AI position will look cheap in retrospect, just as any price paid for early internet infrastructure in 1995 would have been justified.

These worldviews cannot both be correct simultaneously. If AI truly transforms the global economy at the pace Son envisions, Berkshire's cash pile will have been a historic missed opportunity — the equivalent of sitting out the industrial revolution with a vault full of gold coins. Conversely, if AI development hits technical ceilings, regulatory walls, or simply takes decades longer than expected, SoftBank's leveraged bets could unravel spectacularly, echoing the WeWork debacle but at 10 times the scale.

The uncomfortable truth is that we are watching a real-time experiment with hundreds of billions of dollars at stake, and the answer won't be clear for years.

The Market Context: Why This Matters Now

This clash is not happening in a vacuum. The broader AI investment landscape in 2026 provides critical context for understanding both strategies.

• Total global AI investment exceeded $300 billion in 2025 and is projected to reach $500 billion by 2027
• Nvidia's market cap has surpassed $3 trillion, reflecting the market's conviction in AI infrastructure demand
• OpenAI's revenue is reportedly approaching $15 billion annually, up from roughly $4 billion in 2024
• Microsoft, Google, Amazon, and Meta have each committed over $50 billion annually to AI capital expenditure
• AI startup valuations have reached levels not seen since the dot-com era, with pre-revenue companies commanding multi-billion-dollar price tags

In this environment, Berkshire's caution looks increasingly like an outlier. Nearly every major technology company and sovereign wealth fund on earth is placing aggressive AI bets. SoftBank's strategy, while extreme in its concentration, is directionally aligned with the market consensus.

But market consensus has been catastrophically wrong before. The dot-com bubble, the housing crisis, and the crypto collapse all demonstrated that widespread conviction does not equal correctness.

What This Means for Investors and the Industry

For individual and institutional investors, the Berkshire-SoftBank divide offers a useful framework for evaluating their own AI exposure. The key questions are not about specific companies but about fundamental assumptions.

If you believe AI will deliver transformative value within 5-7 years, SoftBank's approach — concentrated bets on leading platforms, tolerance for near-term losses, and aggressive financial engineering — is the rational strategy. The risk is binary: massive upside or devastating loss.

If you believe AI is overhyped, will take decades to mature, or faces significant technical and regulatory headwinds, Berkshire's approach — preserving capital, avoiding speculation, and waiting for the inevitable correction — makes more sense. The risk here is opportunity cost: watching the AI revolution from the sidelines.

For the broader AI industry, these two strategies have practical implications. SoftBank's capital fuels companies like OpenAI, enabling them to pursue expensive research and infrastructure buildouts that might not be possible through organic revenue alone. If SoftBank's model fails, the resulting capital withdrawal could slow AI development significantly.

Looking Ahead: The Verdict May Come Sooner Than Expected

The resolution of this debate may arrive faster than either camp expects. Several catalysts in 2026 and 2027 could tip the scales decisively.

Roze AI's IPO performance will be an early test. If public markets embrace a $100 billion AI holding company with minimal current profits, it validates Son's thesis that investors are willing to fund the AI future on faith. If the IPO stumbles or is delayed, SoftBank's entire financial architecture comes under pressure.

Greg Abel's first major capital allocation decision at Berkshire will signal whether the post-Buffett era brings a philosophical shift. If Abel deploys even a fraction of the $400 billion into technology or AI-adjacent investments, it could represent the most significant strategic pivot in Berkshire's history.

The AI industry itself will provide the ultimate judge. If GPT-5 or equivalent models demonstrate capabilities that clearly justify the trillions being invested — autonomous scientific research, fully reliable coding, or genuine reasoning — the SoftBank thesis wins. If progress plateaus and the 'scaling laws' that have driven recent advances hit diminishing returns, Berkshire's patience will be rewarded.

One thing is certain: by the end of this decade, we will know which philosophy was right. And the financial consequences — for shareholders, for the AI industry, and for the global economy — will be measured in trillions of dollars. The arena in Omaha was half empty. The question is whether that's a sign of decline, or of a world that has simply moved on to a different game entirely.