The initiative was led by Bitcoin advocate Ethan Peck, who argued for using a portion of Meta’s $72 billion cash reserves to hedge against inflation. However, it is believed that Meta CEO Mark Zuckerberg's controlling stake ensured the proposal’s defeat. Despite this, interest in corporate Bitcoin holdings is still rising globally. Meanwhile, Meta is still laser focused on artificial intelligence and defense applications. The company recently partnered with defense firm Anduril to develop AI-driven AR gear for US soldiers, repurposing metaverse technologies for military use. On the other hand, Meta’s AI research also faces internal struggles, with key talent departing and its Llama models trailing competitors. AI chief Yann LeCun is calling for a paradigm shift when it comes to training methods to build more intelligent systems. Bitcoin Proposal Crushed in Meta Vote Meta shareholders decisively voted against a proposal to explore adding Bitcoin to the tech giant’s balance sheet. The “Bitcoin treasury assessment” proposal was submitted earlier this year by Bitcoin advocate Ethan Peck, but it received just 3.92 million votes in favor. This accounts for about 0.08% of the total votes. Nearly 5 billion shareholders rejected the measure, according to a May 28 regulatory filing . It is assumed that Meta CEO Mark Zuckerberg, who controls 61% of the company's voting power, voted against the proposal, which effectively sealed its defeat. (Source: SEC ) Peck’s proposal argued that a portion of Meta’s $72 billion in cash and cash equivalents should be converted into Bitcoin as a hedge against inflation. In his supporting statement, Peck claimed that 28% of Meta’s total assets are “consistently diminishing shareholder value” due to currency debasement and bond yields that fail to outpace inflation. He also pointed to BlackRock’s stance that a 2% allocation into Bitcoin is considered reasonable. Peck is the Bitcoin director at Strive, and submitted the proposal on behalf of his family’s Meta shares. He previously made similar pitches to Microsoft and Amazon. Microsoft shareholders ultimately rejected the idea in December, while Amazon investors are still due to vote on a similar proposal. Despite these high-profile rejections, corporate interest in Bitcoin is still gaining momentum. According to BitcoinTreasuries.NET , 116 publicly traded companies now hold Bitcoin on their balance sheets. Some of the more recent entrants include GameStop and Swedish health tech firm H100, both of which made their first Bitcoin purchases last month. Holders of Bitcoin (Source: BitcoinTreasuries.net ) Strategy is still the largest corporate Bitcoin holder, with 580,250 BTC worth approximately $60.9 billion. Other major holders include Marathon Digital Holdings and Tesla, each with over $1 billion in Bitcoin holdings. For now, the stark rejection at Meta suggests that shareholder appetite for crypto diversification is limited among large-cap tech firms, even as more companies globally are beginning to see Bitcoin as a viable corporate reserve asset. Meta Steps Into Combat While Bitcoin may not be a top priority at Meta, AI certainly is. The social media giant recently partnered with defense contractor Anduril to create artificial intelligence-powered augmented and virtual reality gear for the US military. The main goal of the collaboration is to equip soldiers with advanced headsets that incorporate Anduril’s AI-driven Lattice command and control system. This platform aggregates and interprets data from thousands of sources to deliver real-time battlefield intelligence. According to Anduril, the gear will enhance soldiers' situational awareness and enable more intuitive control over autonomous systems. This will provide a new layer of immersion and decision-making capability during combat missions. The project is privately funded and forms part of the growing trend of repurposing consumer technologies for military applications. Meta has invested $40 billion into virtual and augmented reality technologies since pivoting to its metaverse ambitions in 2021, and is now leveraging its expertise in immersive tech for defense purposes. Anduril’s co-founder Palmer Luckey , who also founded Oculus VR, said the joint project is part of his vision to turn US troops into “technomancers” by using cutting-edge tech to dominate the battlefield. He revealed on X that their first public collaboration will be a system called EagleEye, which is intended to evolve into the Army’s next Soldier Borne Mission Command platform. This new partnership comes after Microsoft initially held a similar AR contract under the same military program using its HoloLens headset. However, Microsoft confirmed in February that Anduril took over the lead, though Microsoft will still provide cloud infrastructure. It is becoming clear that the ties between Silicon Valley and the US defense establishment are deepening. Last November, Meta revised its acceptable use policies to allow its Llama language models to be used in military and national security applications. Around the same time, Anthropic made its Claude AI models available to US defense agencies through Palantir’s AI platform, hosted on AWS. Meta’s AI Vision Diverges On the other hand, Meta’s chief AI scientist Yann LeCun is concerned that current artificial intelligence models fall short of truly replicating human intelligence, and suggested that a fundamental shift in training methods is required to close the gap. At the AI Action Summit in Paris earlier this year, LeCun shared four key traits that define intelligent behavior in animals and humans: understanding the physical world, having persistent memory, reasoning ability, and the capacity for hierarchical planning. He said that today’s large language models, which power popular AI chatbots, lack these essential capabilities and that simply adding new features to existing architectures isn’t enough. Instead, LeCun advocates for a much more sophisticated approach by using what he calls “world-based models,” which will enable AI systems to simulate actions and predict how the world would change as a result. He argues that these models should learn through abstraction, much like humans do, in order to operate in an unpredictable and complex physical environment. Meta started exploring new directions, including retrieval augmented generation (RAG) to enhance model outputs with external data, and its February release of V-JEPA, a non-generative model that learns by predicting missing parts of video sequences. Still, LeCun says the industry needs to move beyond pattern-matching and embrace deeper cognitive modeling. However, Meta is also grappling with internal challenges as its AI division experiences a talent drain . Business Insider reports that only three of the original 14 researchers who developed the company’s first Llama model remain after many moved to Mistral. The exodus took place amid a lukewarm response to Meta’s latest release, Llama 4, which struggled to keep pace with offerings from OpenAI, Google, and Anthropic. These competitors have rolled out models that are praised for more advanced reasoning and user experience. According to The Wall Street Journal, Meta also delayed the release of its next-generation Llama model, code-named “Behemoth.”
