Key Takeaways

• Microsoft’s approximately $13B investment in OpenAI now stands at roughly $228B — delivering a 17x multiple
• SoftBank deployed $64.6B and has unrealized gains exceeding $50B, with holdings valued at approximately $99.3B
• OpenAI CEO Sam Altman maintains no ownership position in the artificial intelligence company
• The nonprofit OpenAI Foundation retains 25.8% ownership acquired at no cost — while maintaining complete board control
• Nvidia’s position shows a slight loss, holding $29.6B in value against a $30.1B investment

An ownership breakdown detailing OpenAI’s shareholder structure surfaced online in early April, sparking widespread discussion about the distribution of value within the AI giant. Investor Sheel Mohnot shared the document on X, which appears compiled from publicly available filings and secondary market transactions. The data reflects OpenAI’s current $852 billion post-money valuation after completing a $122 billion capital raise.

The spreadsheet details each major investor’s ownership percentage, capital deployed, and return multiplier. This information provides unprecedented transparency into the financial structure of one of tech’s most valuable private companies ahead of its anticipated public offering.

Microsoft emerges as the dominant financial beneficiary according to the leaked data. The tech giant deployed approximately $13 billion through multiple funding rounds — beginning with a $1 billion investment in 2019, followed by a massive $10 billion injection in January 2023, plus an additional $2 billion during 2024. The company’s 26.79% ownership stake carries an estimated value of $228.3 billion, representing approximately a 17.6x return multiple. This combination of scale and returns is unmatched among OpenAI’s investor base.

Microsoft’s quarterly SEC filings verify the $13 billion total investment amount. OpenAI recently identified Microsoft as a material business dependency in investor disclosures, referencing revenue-sharing agreements and exclusive cloud infrastructure commitments.

SoftBank stands as the largest capital contributor beyond Microsoft, having pledged $64.6 billion to OpenAI. The Japanese conglomerate’s 11.66% ownership position is currently valued at approximately $99.3 billion. This represents unrealized gains surpassing $50 billion on the investment. CNBC reporting verified that SoftBank completed funding its $40 billion commitment by December 2025, utilizing a $40 billion bridge financing facility arranged by JPMorgan and Goldman Sachs.

Original Backers See Extraordinary Multiples

First movers in OpenAI’s funding history achieved the most impressive return multipliers, despite smaller absolute dollar amounts. Khosla Ventures deployed roughly $50 million during 2019, with that position now valued at approximately $1.5 billion — representing a 30x gain. Sound Ventures, the investment firm co-founded by entertainer Ashton Kutcher, committed between $20–30 million and currently holds about $1.3 billion in value, translating to a 43x multiple. Thrive Capital invested $3.5 billion and maintains a 1.98% stake currently worth $16.9 billion.

Nvidia represents the sole investor showing negative returns. The chipmaker holds 3.47% of OpenAI with a current valuation of $29.6 billion, compared to an initial cost basis of $30.1 billion. Much of Nvidia’s contribution came through GPU compute infrastructure rather than direct cash, making the return calculation more nuanced.

CEO Sam Altman’s Equity Remains Undetermined

Perhaps the most striking revelation in the ownership document is that Sam Altman, who has served as OpenAI CEO since 2019, currently possesses zero equity in the organization. His ownership line item shows as undecided. Board chairman Bret Taylor publicly stated in October 2024 that Altman’s equity arrangement had not been finalized. Altman personally refuted speculation about imminent equity grants during an internal employee gathering.

The original OpenAI Foundation nonprofit entity maintains 25.8% ownership with zero capital invested — theoretically producing an infinite return on a holding valued at approximately $219.8 billion. While representing a minority economic interest, the Foundation retains absolute authority over all board member appointments.

OpenAI is preparing for a public market debut targeted for 2026 or early 2027, with discussions centering on a potential $1 trillion valuation. Industry observers anticipate Altman’s ownership compensation will be addressed in connection with that liquidity event.

Drift Protocol said the April 1 attack on its platform followed months of planning and social engineering. 

The decentralized exchange linked the case to a group that spent time building trust with contributors before sending malicious tools and links. External estimates put the loss at about $280 million.

Drift Protocol said its early review found a long and organized campaign against the platform. The team said the attackers showed “organizational backing, resources, and months of deliberate preparation” during the operation.

The exchange said the contact began around October 2025. According to Drift, people posing as members of a quantitative trading firm approached contributors at a major crypto conference and claimed they wanted to integrate with the protocol.

Drift said the group kept meeting contributors at several industry events over the next six months. The team said the people involved were technically skilled, knew how Drift worked, and appeared to have real professional backgrounds.

That steady contact helped the group gain trust. Drift said the attackers later used malicious links and tools shared with contributors to compromise devices, carry out the exploit, and remove traces of their activity after the breach.

In addition, Drift said it has “medium-high confidence” that the same actors behind the October 2024 Radiant Capital hack carried out this exploit. That earlier attack caused losses of about $58 million and also involved malware used to gain access to internal systems.

Radiant Capital said in December 2024 that a North Korea-aligned hacker posed as a former contractor and sent malware through Telegram. Radiant said “this ZIP file” later spread among developers for feedback and opened the way for the intrusion.

Drift warns conferences can become attack targets

Drift said the people who met contributors in person “were not North Korean nationals.” At the same time, the team said DPRK-linked threat actors often use third-party intermediaries for face-to-face contact and relationship building.

The exchange said it is now working with law enforcement and other crypto industry participants to build a full record of the April 1 attack. 

The case has also added a fresh warning for crypto firms, as conferences and in-person meetings can give threat groups a chance to study teams, build trust, and prepare later attacks.

Key Highlights

• Marvell’s shares climbed 21.3% throughout March, fueled by robust quarterly results and a transformative Nvidia partnership
• Fourth-quarter revenue increased 22.1% to $2.2 billion, while adjusted earnings per share rose 33.3% to $0.80
• Nvidia committed $2 billion to Marvell through an equity investment and unveiled a comprehensive strategic collaboration
• Company executives project 40% data center revenue expansion in fiscal 2027, significantly exceeding Wall Street’s 25% forecast
• Erste Group launched coverage with a Buy recommendation on April 2, highlighting robust financial metrics and AI market positioning

March proved to be a landmark month for Marvell Technology. The semiconductor company not only exceeded expectations with its quarterly performance but also secured a game-changing agreement with Nvidia.

Marvell Technology, Inc., MRVL

The fourth-quarter financial results impressed investors. Revenue surged 22.1% compared to the prior year, reaching $2.2 billion. Adjusted earnings per share hit $0.80, representing a 33.3% increase. Both metrics exceeded Wall Street projections.

Forward-looking guidance reinforced the positive momentum. Executives forecast a 9% sequential revenue boost in the first quarter, with adjusted EPS projected at $0.79. These projections also surpassed analyst expectations.

However, the month’s most significant development arrived on March 31. Nvidia revealed a $2 billion equity stake in Marvell, accompanied by an extensive strategic alliance.

The collaboration encompasses custom silicon development, networking solutions, and optical technology innovations. At its core sits NVLink Fusion, Nvidia’s framework for incorporating external chips into its artificial intelligence infrastructure ecosystem.

The arrangement’s significance lies in the architectural flexibility it creates. Previously, AI infrastructure typically followed two paths: Nvidia-centric systems or custom XPU chip configurations with Ethernet connectivity. This alliance introduces hybrid possibilities — combining XPUs with Nvidia’s GPUs, CPUs, and interconnect technologies.

Data Center Revenue Projections Exceed Expectations

Management established ambitious targets for fiscal 2027. The chipmaker anticipates data center revenue will expand by 40% — substantially surpassing the 25% consensus among financial analysts.

This optimism appears rooted in its XPU operations, which supply custom AI chip intellectual property to cloud computing giants. While concerns had emerged about potential market share erosion at Amazon following the launch of Amazon’s Trainium chips, the aggressive guidance indicates the XPU business remains robust.

Marvell has simultaneously diversified its client portfolio. Microsoft introduced its enhanced Maia2 XPU processor in January, incorporating Marvell’s intellectual property into the chip architecture.

The Nvidia alliance also encompasses silicon photonics — an emerging technology poised to eventually supplant copper-based networking within AI data centers. Given that Nvidia’s existing NVLink infrastructure relies on copper, the partnership with Marvell represents a strategic shift toward optical interconnect solutions.

Wall Street Coverage Intensifies

On April 2, Erste Group launched coverage of Marvell with a Buy recommendation. The investment firm noted that net profits have doubled across the previous five quarters, while return on equity has climbed to 19%.

Erste additionally emphasized Marvell’s competitive advantages in high-performance analog and optical DSP technologies as fundamental drivers behind its optimistic assessment.

The Nvidia investment propelled Marvell to 52-week peak valuations. The stock had traded within a constrained range for much of the preceding six months, but the convergence of impressive earnings and Nvidia’s endorsement triggered a breakout.

Marvell presently trades at approximately 27 times forward earnings projections — a valuation premium compared to last year, though one that numerous analysts consider justified given the data center expansion outlook.

The company’s XPU offerings now interface with Nvidia’s NVLink Fusion infrastructure, potentially unlocking additional revenue opportunities throughout Nvidia’s expanding network of hyperscale clients.

Anthropic, the AI safety-focused lab behind several widely used language models, has moved to formalize its political engagement by launching an employee-funded political action committee named AnthroPAC. A filing with the Federal Election Commission shows the organization as a connected entity to Anthropic, organized as a separate segregated fund and aimed at receiving voluntary contributions from employees. The filing outlines the PAC’s intent to participate in federal elections while remaining aligned with the company’s stated interest in AI policy and safety considerations.

Under U.S. campaign finance rules, individual contributions to a federal candidate are capped at $5,000 per election, with disclosures required through public filings. AnthroPAC’s organizers say the fund is designed to support candidates from both major parties. However, observers and industry watchers are already raising questions about how closely the effort will stay within bipartisan lines, given broader debates over AI regulation, safety standards, and the strategic direction of AI policy in Washington.

The AnthroPAC move lands as Anthropic navigates a fraught relationship with the U.S. government over how its technology should be employed. Separately, the Defense Department in February designated Anthropic as a supply chain risk—an action tied to the company’s stance against the use of its AI in fully autonomous weapons and mass surveillance. Anthropic has challenged that designation in court, contending it constitutes retaliation for a protected position. A federal judge in California has temporarily blocked the measure and paused further restrictions while the dispute unfolds.

Beyond governance and defense concerns, Anthropic has already been active politically this cycle. Notably, the company contributed $20 million to Public First Action, a political committee focused on AI safety and related policy advocacy, underscoring the firm’s broader strategy to influence AI-related regulation and public safety standards.

Meanwhile, Anthropic’s broader ecosystem is drawing capital and infrastructure support that could accelerate its technology roadmap. In a related development, Google is preparing to back a multibillion-dollar data-center project in Texas that would be leased to Anthropic via Nexus Data Centers. The project’s initial phase could exceed $5 billion, with Google expected to provide construction loans and be joined by banks arranging additional financing. The arrangement highlights the growing demand for AI infrastructure capable of supporting expansion in model training, inference, and data storage.

Key takeaways

• Anthropic formed AnthroPAC, an employee-funded political action committee registered as a separate segregated fund under the company’s umbrella.
• The PAC is intended to support candidates from both parties, with strict contribution limits and mandatory disclosures under U.S. election law.
• The move occurs amid fraught relations with the Pentagon over AI use, including a safety-focused designation that Anthropic is challenging in court.
• Anthropic has a track record of political giving in this cycle, including a $20 million contribution to Public First Action focused on AI safety.
• Google’s backing of a Texas data-center project for Anthropic signals strong infrastructure demand and potential financing mechanisms that could accelerate AI deployment.

Anthropic’s political engagement and the policy context

The formation of AnthroPAC marks a notable step in how AI firms engage with lawmakers and regulators. By coordinating staff contributions through a dedicated PAC, Anthropic signals a structured approach to influencing elections and policy debates that shape the development and governance of artificial intelligence. The FEC filing describes AnthroPAC as a “connected organization” operating under a separate segregated fund, aligning with typical industry practices for corporate-employee political activity. While the stated aim is bipartisanship, the broader AI policy environment in the United States has become highly polarized, with differing views on liability, safety mandates, data privacy, and government access to AI systems.

Investors and builders watching the space can interpret this as part of a broader trend: major AI developers increasingly engage directly in policy conversations, seeking to frame the regulatory environment in ways that balance innovation with oversight. The implications extend beyond ethics and governance; policy direction can materially affect the regulatory runway for product development, procurement, and collaboration with public sector actors. The presence of a formal PAC also raises questions about how corporate political contributions could influence which AI-safety and governance proposals gain traction on Capitol Hill and in regulatory agencies.

Defense frictions and legal maneuvering

The tension between Anthropic and the Department of Defense centers on how the company’s models should be deployed in sensitive contexts. The Pentagon’s decision to label Anthropic as a supply chain risk stemmed from the company’s public stance against fully autonomous weapons and broad surveillance use. Anthropic has challenged that designation in court, arguing that it amounts to retaliation for a viewpoint it regards as legitimate and protected. A federal judge in California issued a temporary ruling to pause the measure and related restrictions while the case proceeds, illustrating the jurisdictional balance between corporate risk assessments and national-security considerations in AI technology usage.

For policymakers, the case underscores a core policy question: where should the line be drawn between compelling safety and preserving innovation? If courts narrow how procurement risk designations can be wielded, it could affect how similar technology providers are treated as the government expands its AI procurement and testing programs. Conversely, if the government can justify risk designations on safety grounds, it could strengthen leverage for tighter controls on how AI systems are used in defense contexts.

Political giving and AI-safety advocacy

Anthropic’s political activity isn’t limited to its new PAC. Earlier in the cycle, the company contributed a sizable $20 million to Public First Action, a political arm focused on AI safety and public-interest considerations tied to the development and governance of AI technologies. This level of funding signals a broader strategy to influence public discourse and regulatory design around AI, complementing the PAC’s electoral role with policy advocacy and education efforts. Observers are watching how such funding patterns translate into concrete policy outcomes, particularly in an environment where legislators are weighing landmark AI bills and safety standards that could shape model development, data usage, and transparency requirements.

Infrastructure bets amid AI acceleration

Infrastructure matters are increasingly central to AI strategy, and Google’s involvement in a Texas data-center project for Anthropic is a vivid illustration. The Nexus Data Centers-leased facility, if realized as outlined, could become a cornerstone asset to support large-scale model training and deployment. The project’s initial phase exceeding $5 billion underscores the capital intensity of modern AI initiatives and the financial orchestration that underpins them. Google’s expected role in providing construction loans, alongside competitive financing arrangements from banks, points to the consolidation of AI infrastructure finance as a distinct sub-market within the tech sector. For Anthropic and similar firms, such backing could shorten timelines to deploy more capable models and scale services that demand robust, energy-efficient, and highly reliable data-center capacity.

As policy debates progress, industry participants and investors should monitor both political and practical developments: how much traction new AI safety proposals gain in Congress, how procurement rules evolve in defense programs, and how infrastructure financing evolves to accommodate the next wave of AI workloads. Each of these strands will influence not only which AI products reach market first, but also how quickly the industry can translate research advances into real-world use cases across enterprise, healthcare, and public services.

Readers should stay attentive to any updates on Anthropic’s PAC activity and the Pentagon case outcomes, as both arenas will shape the company’s public-facing strategy and its broader partnerships. The balance between safety-driven governance and aggressive innovation remains a live tension set to define the next phase of AI adoption and investment.

The rapid growth in the number of crypto tokens is outpacing the value they generate, creating an “existential” problem for the industry, according to Michael Ippolito, co-founder of Blockworks.

In a series of posts on X, Ippolito noted that while total crypto market capitalization remains relatively strong, the average value per token tells a different story. “The average coin is only slightly higher than where it was in 2020 (!) and down ~50% since 2021,” he wrote.

Median token returns have also deteriorated sharply. Most tokens are down roughly 80% from their highs, suggesting that gains have been concentrated in a narrow set of large-cap assets, while the broader market underperforms, Ippolito claimed.

He argued that the imbalance appears to be driven by a rapid expansion in token supply. “We created a TON of new assets and STILL total market cap is flat,” he wrote, adding that this dynamic effectively dilutes value across a growing pool of tokens.

Related: Bitcoin ‘done’ with 85% crashes, says Cathie Wood amid new $34K target

Token prices break from fundamentals

Ippolito also claimed that the relationship between fundamentals and price has weakened. In 2021, token prices closely tracked onchain revenue. Recent data shows that despite a resurgence in protocol revenues, prices have not followed, pointing to a disconnect between usage and investor returns.

He argued that this signals a loss of confidence in tokens as vehicles for capturing value. “The token problem is existential for this industry,” he said, adding that without stronger alignment between fundamentals and price, the sector risks losing its core appeal.

In a post on X, Arthur Cheong, founder and CEO of DeFiance Capital, said he agrees “with the urgency to fix the current situation of tokens in the crypto industry,” warning that if the market continues to concentrate around a small set of assets like Bitcoin and Ether, the broader crypto ecosystem risks losing relevance.

Related: Bitcoin shorts risk $2.5 billion liquidation at $72K: Are bears in danger?

Capital shifts from tokens to stocks

Investor demand is increasingly moving away from newly launched tokens toward publicly listed crypto firms, as most token launches fail to hold value, a February research from DWF Labs found. The report revealed that over 80% of projects trade below their token generation event (TGE) price, with typical losses of 50% to 70% within about three months.

The pattern appears structural rather than cyclical. According to DWF’s Andrei Grachev, most tokens peak within the first month before declining under sustained selling pressure. Factors such as airdrops and early investor unlocks add to the supply overhang, reinforcing downward price trends even for projects with active products or protocols.

Magazine: Bitcoin may take 7 years to upgrade to post-quantum — BIP-360 co-author