Introduction: Why Companies Sometimes Sleep with the Frenemy
The AI era is reshaping the foundational logic of corporate rivalry. For decades, the dominant grammar of business strategy was adversarial: outbuild, outspend, outmaneuver. Competition was not merely an operating condition — it was a corporate identity. Companies defined themselves against their adversaries. Gates versus Jobs. Google versus Microsoft. Facebook versus Twitter. The drama of technology capitalism was inseparable from its combat.
But artificial intelligence has introduced a new and deeply inconvenient variable into that grammar: infrastructure costs have become so astronomically large that even the most entrenched of rivals may find cooperation not merely rational — but existentially necessary. This is not friendship. This is not ideological truce. This is strategic coexistence — the uncomfortable calculus of two players who hold each other in contempt and need each other more than they need the satisfaction of victory.
This paper introduces the “Sleeping with the Frenemy” framework — a model describing the structural conditions under which direct competitors, ideological antagonists, and public adversaries temporarily align in cooperation when capital intensity, infrastructure bottlenecks, survival pressures, or time-to-market economics make that alignment superior to sustained rivalry. The AI economy is peculiarly — perhaps uniquely — susceptible to this phenomenon. Compute is expensive. Energy is scarce. Advanced semiconductors are constrained by geopolitical supply chains. Datacenter construction timelines extend across years. Regulatory approvals for power infrastructure take half a decade. And in the race for artificial superintelligence, hyperscale growth rewards the fastest, not the proudest.
The International Monetary Fund, in its landmark World Economic Outlook of April 2025, identified AI as a once-in-a-generation force capable of raising the average pace of annual global economic growth. The IMF’s working paper Power Hungry: How AI Will Drive Energy Demand — authored by economists Bogmans, Ganpurev, Gomez-Gonzalez, Melina, Pescatori, and Thube — found that AI-producing sectors in the United States have grown at nearly triple the rate of the private non-farm business sector, and that electricity costs for vertically integrated AI companies nearly doubled between 2019 and 2023.1 The paper concluded with a warning of historical gravity:
— IMF Working Paper WP/25/81, Bogmans et al. (April 2025):
“The AI boom will cause manageable but varying increases in energy prices and emissions depending on policies and infrastructure constraints. Under scenarios with constrained growth in renewable energy capacity and limited expansion of transmission infrastructure, U.S. electricity prices could increase by 8.6%.”
It is against this backdrop of civilizational capital deployment that the central case of this paper must be understood. Meta Platforms — the social media colossus controlled by Mark Zuckerberg — has entered into a $1.2 billion energy infrastructure arrangement for its Cowboy Project in Wyoming, with Tesla supplying the Megapack batteries that will store and dispatch power to Meta’s AI data center operations.2 The two companies are not allies. Their chief executives spent nearly a decade trading philosophical broadsides, performing corporate hostility, and even challenging each other to physical combat. Yet in May 2026, they are bound together by the oldest of economic forces: the imperative to keep the machines running.
In Q1 2026, Meta CEO Mark Zuckerberg opened his earnings call by declaring:
— Mark Zuckerberg, Meta Q1 2026 Earnings Call (April 29, 2026):
“We had a milestone quarter with strong momentum across our apps and the release of our first model from Meta Superintelligence Labs. We’re on track to deliver personal superintelligence to billions of people.”
On the same call, CFO Susan Li warned analysts that Meta could keep underestimating its compute needs as AI model adoption accelerates.3 Meanwhile, on Tesla’s Q1 2026 earnings call on April 22, 2026, CEO Elon Musk stated:
— Elon Musk, Tesla Q1 2026 Earnings Call (April 22, 2026):
“We’ve got a very exciting year ahead of us with 2026. We’re going to be substantially increasing our investments in the future, so you should expect to see a very significant increase in capital expenditures. I think it’s well justified for a substantially increased future revenue stream. Obviously, Tesla is not alone in this. You’ve seen most, if not all, certainly the major technology companies substantially increasing their capital investments.”
Two companies. Two rival empires. One physical constraint. This paper traces the full anatomy of that convergence.
Section 1: The Frenemy Playbook — A History of Tech Rivals Becoming Partners
Before turning to Musk and Zuckerberg, it is essential to establish that their arrangement is not historically unprecedented. The history of modern technology is littered with moments when yesterday’s adversary became today’s indispensable partner. What artificial intelligence has done is accelerate the frequency and scale of these moments, transforming episodic exceptions into structural patterns. Professors Marco Iansiti and Karim R. Lakhani of Harvard Business School — in their landmark work Competing in the Age of AI: Strategy and Leadership When Algorithms and Networks Run the World (Harvard Business Review Press, 2020) — warned precisely of this collapse of traditional industry lanes, arguing that AI-driven firms redefine competitive boundaries entirely:
— Professors Marco Iansiti & Karim R. Lakhani, Harvard Business School, Competing in the Age of AI (2020):
“AI-centric organizations exhibit a new operating architecture, redefining how they create, capture, share, and deliver value. AI-driven processes are vastly more scalable than traditional processes, drive massive scope increase enabling companies to straddle industry boundaries, and enable powerful opportunities for learning.”
The strategic implication is clear: when AI obliterates industry boundaries, competitive logic itself must be renegotiated. Companies that once occupied separate lanes now compete — and cooperate — across the same territory simultaneously.
1.1 Apple + Microsoft: Saving the Enemy (1997)
On August 6, 1997, at the Macworld Boston conference, Steve Jobs made what Apple devotees considered an act of corporate heresy: he announced a $150 million investment from Microsoft — Apple’s most hated rival — at a moment when Apple was losing $1 billion annually and had fewer than ninety days of liquidity remaining.4 When Bill Gates appeared via satellite on the conference screen, the audience booed. But behind the theatre was a transaction of extraordinary strategic intelligence: Microsoft received nonvoting shares, antitrust insurance, and a guaranteed five-year commitment to Office for Mac; Apple received capital, credibility, and survival.
— Steve Jobs, D5 Technology Conference (2007):
“Apple didn’t have to beat Microsoft. Apple had to remember who Apple was because they’d forgotten who Apple was.”
The strategic lesson is the first pillar of the Sleeping with the Frenemy playbook: sometimes preserving a rival is smarter than eliminating them. Microsoft ultimately sold its Apple stake in 2003 for $550 million — a return of roughly 260 percent over six years.5
1.2 Apple + IBM: Enterprise Reinvention (2014)
In 2014, Apple and IBM — companies whose rivalry stretched back to Apple’s iconic 1984 Super Bowl advertisement depicting IBM as a totalitarian overlord — announced a sweeping enterprise partnership. Apple brought consumer hardware excellence and the most loyal user base in the history of personal technology; IBM brought Watson AI infrastructure and penetration into exactly those corporate accounts Apple had never cracked. The deal illustrated a second key principle: competitors partner when their strengths are asymmetric. The very differences that made them rivals in one domain made them indispensable to each other in another.
1.3 Apple + Google: From Platform War to AI Cooperation (January 2026)
Perhaps the most striking contemporary example of frenemy capitalism prior to the Meta-Tesla arrangement arrived in January 2026: Apple and Google announced a multiyear partnership to integrate Google’s Gemini AI models into Apple Intelligence, powering a rebuilt Siri that Apple had been unable to deliver through internal development alone.6 The context is saturated with irony. Apple and Google had fought an existential platform war across smartphones for nearly two decades. Yet Google had simultaneously been paying Apple an estimated $20 billion per year as the default search engine on iOS.7
The deal, reportedly valued at approximately $1 billion annually, was described by Dan Ives, Senior Equity Analyst at Wedbush Securities, as:
— Dan Ives, Senior Equity Analyst, Wedbush Securities (January 2026):
“A stepping stone to accelerate its AI strategy into 2026 and beyond.”
Professor Rita McGrath of Columbia Business School — ranked #1 for Strategy by Thinkers50 and a leading voice on competitive dynamics — addressed the broader structural context at the HBR Strategy Summit 2026, arguing that in the AI era, companies can no longer depend on long-term competitive advantage:
— Professor Rita McGrath, Columbia Business School, HBR Strategy Summit 2026 (February 2026):
“This shift means companies can no longer depend on long-term competitive advantage. Instead, in an AI world, companies need to choose a ‘center’ around which to build capacity and allocate resources.”
The Apple-Google deal is a case study in exactly that logic: two companies with irreconcilable platform ambitions nonetheless cooperating because the AI infrastructure gap is wider than the competitive gap.
1.4 The Broader Coopetition Pattern: OECD, AEI, and Academic Consensus (2024–2026)
The OECD, in its November 2025 paper Competition in Artificial Intelligence Infrastructure, documented how AI supply chains have created structural dependencies that make coopetition not merely common but architecturally necessary: companies simultaneously compete in AI applications while depending on each other — and on shared suppliers like Nvidia — for the physical infrastructure that makes those applications possible.8
Yale School of Management Professors Edward Snyder and Logan Bender, writing in Yale Insights (October 2025), encapsulated the competitive collapse of traditional industry categories with arresting precision:
— Professors Edward A. Snyder & Logan Bender, Yale School of Management, Yale Insights (October 2025):
“The Magnificent Seven — Alphabet, Amazon, Apple, Microsoft, Meta, Nvidia, and Tesla — used to compete in distinct lanes. AI has changed that. These seven are now in a new pool without lanes and with more competitors. Over the history of capitalism, no other new technology has attracted as much aggregate investment as AI.”
The World Bank’s Digital Progress and Trends Report 2025: Strengthening AI Foundations (November 2025) added an equity dimension, warning that high-income countries account for 87 percent of notable AI models, 86 percent of AI startups, and 91 percent of venture capital funding — despite representing only 17 percent of the global population.9 The concentration of AI infrastructure is not merely a competitive dynamic within the tech industry — it is reshaping the geopolitical distribution of economic power globally.
Section 2: Musk vs. Zuckerberg — Anatomy of a Modern Tech Rivalry
The historical arc of tech cooperation demonstrates that frenemy capitalism is not new. But the specific rivalry between Elon Musk and Mark Zuckerberg represents the genre at its most extreme, most public, and most psychologically revealing. These are not merely two executives who disagree about business strategy. They are two individuals who have expressed, at various points, near-total contempt for each other’s worldview, intelligence, character, and purpose. Understanding the depth of their antagonism is essential to appreciating the full audacity of what has now occurred between their companies.
2.1 Phase One: Infrastructure Collision (September 1, 2016)
The first visible fracture arrived not through words but through fire. On September 1, 2016, at 9:07 in the morning Eastern Time, a SpaceX Falcon 9 rocket exploded on the launch pad at Cape Canaveral during a routine pre-launch static fire test. The blast destroyed not only the rocket but its payload: the AMOS-6 communications satellite, which Facebook had co-leased with French operator Eutelsat at a cost of approximately $95 million over five years.10 The satellite was central to Facebook’s Internet.org initiative to provide broadband connectivity to sub-Saharan Africa. Zuckerberg was in Kenya when he learned of the explosion and posted:
— Mark Zuckerberg, Facebook post (September 1, 2016):
“As I’m here in Africa, I’m deeply disappointed to hear that SpaceX’s launch failure destroyed our satellite that would have provided connectivity to so many entrepreneurs and everyone else across the continent.”
Musk offered no public response. The silence was its own kind of message. The structural lesson — that infrastructure conflict between rivals generates direct financial damage and lasting personal wounds — would be borne out repeatedly in the years ahead.
2.2 Phase Two: The Philosophical War Over Artificial Intelligence (July 2017)
In July 2017, during a casual Facebook Live broadcast from his Palo Alto backyard, Zuckerberg was asked about Musk’s increasingly alarming warnings about AI’s existential risks to humanity. Musk had been telling audiences of governors and policymakers that artificial intelligence represented a “fundamental risk to the existence of human civilization.” Zuckerberg was dismissive, calling AI doomsday warnings “pretty irresponsible,” and arguing that people who think negatively about AI cannot explain “in good conscience” why they would argue against technologies that could save lives through better medicine and safer vehicles.
Musk read the coverage and responded with a tweet that became one of the most quoted broadsides in the history of Silicon Valley:
— Elon Musk, Twitter (July 25, 2017):
“I’ve talked to Mark about this. His understanding of the subject is limited.”
The tweet was not a broad personal attack. It was a surgical strike on intellectual credibility — Zuckerberg’s most precious social currency. The philosophical divide was now fully public: Musk viewed AI as a civilizational risk requiring radical precaution; Zuckerberg viewed it as a civilizational opportunity requiring radical acceleration. These were not different opinions. They were different cosmologies.11
2.3 Phase Three: Platform Hostility — Cambridge Analytica and the #DeleteFacebook Era (2018)
In March 2018, the Cambridge Analytica scandal revealed that Facebook had allowed the data of tens of millions of users to be harvested without consent and used by a political consultancy. The #DeleteFacebook movement erupted with institutional support. Brian Acton, the WhatsApp co-founder who had sold his company to Facebook for $19 billion, posted simply: “It is time. #deletefacebook.” Musk, characteristically, did not merely comment — he deleted both Tesla’s and SpaceX’s official Facebook pages, each with millions of followers, performing competitive rivalry as moral positioning. Competition had become tribal.
2.4 Phase Four: The Twitter Acquisition and the Birth of Threads (2022–2023)
In October 2022, Elon Musk completed his $44 billion acquisition of Twitter — one of the most chaotic corporate transactions in social media history.12 Musk reimagined the platform as the foundation of a payments, communications, and media empire; he renamed it X and set about transforming its content moderation, verification systems, and editorial philosophy. The chaos of that transformation created precisely the platform vulnerability that Zuckerberg had been waiting to exploit.
Meta launched Threads in July 2023. Within a single week, it had accumulated 100 million users — a growth velocity that no social platform had previously achieved.13 Musk responded with the challenge that briefly captured global attention, issued on Twitter/X in June 2023:
— Elon Musk, Twitter/X (June 2023):
“I’m willing to have a cage match with Zuckerberg if he is.”
Zuckerberg, who had spent the COVID-19 pandemic training seriously in Brazilian jiu-jitsu, replied with three words quoting the legendary fighter Khabib Nurmagomedov: “Send me location.”
2.5 Phase Five: Spectacle Capitalism — The Cage Match That Never Was (2023)
What followed was months of performed rivalry — both men training publicly, Dana White of the UFC declaring both fighters “dead serious,” global media treating the confrontation as both absurdity and spectacle. In August 2023, Zuckerberg finally signaled closure:
— Mark Zuckerberg, Threads (August 13, 2023):
“If Elon ever gets serious about a real date and official event, he knows how to reach me. Otherwise, time to move on. I’m going to focus on competing with people who take the sport seriously.”
Dr. Mark Travers, writing in Psychology Today, diagnosed the episode with precision that extends beyond psychology into media economics:
— Dr. Mark Travers, Psychology Today (September 13, 2023):
“The Musk-Zuckerberg faceoff reflects deep societal shifts. Digital interconnectivity, real-time reactions and polarizing ideologies converge, challenging our traditional notions of rivalry. Beyond boardroom wars, and beyond a cage match, the widespread interest in this fight is a reflection of our complex digital age, and our ever-evolving appetite for spectacle and narrative.”
Modern CEOs have learned that weaponizing public narrative is itself a form of competitive strategy. The fight that never happened was, in its way, more strategically consequential than any fight that did.14
2.6 Phase Six: Divergent Empires, Convergent Economics (2024–2026)
By early 2026, both men had constructed empires of extraordinary scope. Musk’s portfolio spanned Tesla’s automotive and energy divisions, SpaceX’s launch and Starlink connectivity operations, xAI’s Grok language model, the X platform, and Neuralink’s brain-computer interface ambitions. Zuckerberg’s Meta encompassed Facebook, Instagram, WhatsApp, the Llama family of open-source AI models, the newly launched Meta Superintelligence Labs, and one of the most aggressive data center buildout programs in corporate history.
In Q1 2026, Meta reported capital expenditures of $19.84 billion for the single quarter and raised its full-year 2026 capex guidance to between $125 billion and $145 billion — up from a previous range of $115 billion to $135 billion.15 To put that in historical proportion: Meta is now guiding to nearly double what it spent in 2025, and more than it spent in 2024 and 2025 combined — in a single year. The company cited higher component pricing and “additional data center costs to support future year capacity.”
Tesla’s Q1 2026 earnings, reported on April 22, 2026, revealed a company that had fundamentally repositioned itself. Tesla’s energy generation and storage segment posted a gross margin of 39.5 percent — more than double the 17 percent automotive margin in the same period.16 CFO Vaibhav Taneja confirmed on the earnings call that Tesla’s full-year 2026 capex would exceed $25 billion, covering AI compute expansion, Megapack 3 production ramp, new factory construction, and the Cybercab and Optimus robotics programs.17
Two rival empires. Both consuming electrons at civilizational scale. Both constrained by the same physical reality: power supply. The convergence was not ideological. It was electrical.
Section 3: The Meta–Tesla Energy Alliance — Sleeping with the Frenemy in Practice
When the Cowboy Project was announced in May 2026, it generated headlines not merely because of the participants’ identities — though the Musk-Zuckerberg irony was widely noted — but because of its scale, its structural sophistication, and what it revealed about the infrastructure economics of artificial intelligence at civilizational scale. To understand its full significance requires examining each of its components with precision.
3.1 The Cowboy Project: Architecture of a $1.2 Billion Necessity
The Cowboy Project is a solar-plus-storage facility under development near Cheyenne, Wyoming, representing the first phase of a broader energy partnership between Meta Platforms and Canadian energy infrastructure giant Enbridge. Phase One consists of 365 megawatts of solar generation capacity paired with a 200 MW / 1,600 MWh battery energy storage system — one of the largest utility-scale battery installations in the United States dedicated to data center operations.18
Enbridge will invest approximately $1.2 billion in construction, with the project targeted to enter commercial service by the end of 2027. Power will be delivered to Meta through Cheyenne Light, Fuel and Power under Wyoming’s Large Power Contract Service tariff — a mechanism designed specifically to serve large-load customers such as data centers without imposing cost increases on residential ratepayers. The battery component is governed by a long-term battery tolling agreement. And the supplier of those batteries — the entity responsible for manufacturing, delivering, and servicing the 1,600 MWh of electrochemical storage — is Tesla.19
Allen Capps, Enbridge’s Senior Vice President of Corporate Strategy and President of Power, articulated the strategic significance of the arrangement:
— Allen Capps, Enbridge Senior VP Corporate Strategy & President of Power (May 2026):
“The project will be one of the larger utility-scale battery installations supporting U.S. data center operations and growth. Combining utility-scale solar with battery storage delivers reliable, scalable energy solutions for Meta’s data centres and enhances grid reliability for the broader community.”
Meta’s Head of Clean and Renewable Energy, Amanda Yang, elaborated on the dual mandate of the project — not merely powering Meta’s own operations but contributing measurably to grid stability for the broader Wyoming region:
— Amanda Yang, Meta Head of Clean and Renewable Energy (May 2026):
“We’re committed to supporting projects that add new energy to the grid while strengthening reliability in our data center communities. In partnership with Enbridge and Cheyenne Light, Fuel and Power, the Cowboy Project’s 1,600 MWh battery system paired with 365 MW of solar, will deliver flexible, reliable power that benefits the broader grid, including our data center operations.”
The Cowboy Project is not an isolated transaction. It is the latest phase of a growing partnership between Enbridge and Meta that now encompasses approximately 1.6 gigawatts of contracted renewable capacity across North America, including the Clear Fork Solar project (600 MW), the Easter Wind project (152 MW), and the Cone Wind project (300 MW) — all in Texas.20 Enbridge CEO Greg Ebel has disclosed that the company is actively pursuing more than 50 data center-related energy projects across the continent.
3.2 Why Tesla Matters: The Strategic Value of the Megapack
The presence of Tesla in the Cowboy Project reflects Tesla’s emergence as the dominant supplier of utility-scale battery storage in the United States — and, by extension, the dominant provider of the infrastructure that makes variable renewable energy compatible with the relentless, non-negotiable power demands of AI computation. Tesla’s Megapack product line — industrial-scale lithium iron phosphate battery systems — has become the reference product for utility-scale storage the way the iPhone became the reference product for smartphones.
Tesla’s Megapack 3, launched in September 2025 at the RE+ trade show under the Megablock configuration, achieves a site-level energy density of 248 MWh AC per acre — extraordinary compression for constrained sites.21 Tesla’s energy generation and storage segment reached a gross margin of 39.5 percent in Q1 2026, compared to just 17 percent for automotive in the same quarter. The energy business, with $12.8 billion in annual revenue and a project pipeline exceeding $29 billion, has effectively ceased to be a subsidiary of an automotive company and become a standalone infrastructure enterprise with implications that extend far beyond vehicle manufacturing.22
For Meta, Tesla’s Megapack solves a problem that cannot be solved through software optimization or algorithmic ingenuity: the physical reality that solar panels produce no electricity after sunset, and that AI datacenters require constant, uninterrupted power twenty-four hours a day. Battery storage is the bridge between the intermittency of renewable generation and the continuity demands of AI computation. Without it, solar-powered AI infrastructure is a theoretical aspiration rather than an operational reality.
3.3 The Economics of AI Energy: When Utilities Become Strategic Assets
The scale of the energy challenge confronting the AI industry is without modern precedent. In April 2026, the International Energy Agency published its landmark report Key Questions on Energy and AI, finding that global electricity demand from data centers had surged by 17 percent in 2025 alone — vastly outstripping the 3 percent growth in total global electricity demand. Electricity consumption from AI-focused data centers grew even faster, surging 50 percent in a single year.23
The IEA projects that electricity consumption from data centers will roughly double from 485 TWh in 2025 to approximately 950 TWh by 2030 — accounting for nearly 3 percent of total global electricity demand. AI-focused data center electricity consumption is projected to triple over the same period. The IEA further observed, with remarkable force, that the capital expenditure of just five major technology companies now exceeds global investment in oil and natural gas production combined.24
The IMF’s World Economic Outlook (April 2025) contextualized this energy demand within a framework of macro-economic risk, noting that AI’s electricity demands have transformative implications for global economic growth:
— IMF Blog, Ganchimeg Ganpurev & Andrea Pescatori, Research Department (May 13, 2025):
“AI has the potential to raise the average pace of annual global economic growth. Despite challenges related to higher electricity prices and greenhouse gas emissions, the gains to global GDP from AI are likely to outweigh the cost of the additional emissions. The social cost of these extra emissions is minor compared with the expected economic gains from AI, yet it still adds to the worrisome buildup of emissions.”
The Brookings Institution, in its April 2026 background briefing for the Forum for Cooperation on AI, offered the most vivid scalar comparison yet: if data centers were a sovereign nation, their energy consumption would place them among the five largest energy consumers on earth — between Japan and Russia.25 In the United States, data centers are projected to account for nearly half of all electricity demand growth between now and 2030.
The UNDP’s December 2025 flagship report The Next Great Divergence: Why AI May Widen Inequality Between Countries positioned this energy and compute concentration as a global equity crisis in the making. Philip Schellekens, Chief Economist for UNDP Asia Pacific, warned at a Geneva press briefing:
— Philip Schellekens, Chief Economist, UNDP Asia Pacific Regional Bureau (December 2025):
“We think that AI is heralding a new era of rising inequality between countries, following years of convergence in the last 50 years. If inequality continues to rise, the spillover effects in terms of the security agenda, in terms of undocumented forms of migration, will also become more daunting.”
The UNDP report noted that AI had reached 1.2 billion users in only three years, with nearly 70 percent of them in developing countries — yet in many low-income countries, AI usage remains close to 5 percent, while in some high-income economies, two in three people already use AI tools regularly.26 The infrastructure deals being struck between Meta and Tesla in Wyoming are, from this perspective, not merely corporate transactions. They are concentrations of civilizational capability in the world’s wealthiest infrastructure ecosystems — with enormous implications for who benefits from the AI revolution and who does not.
3.4 From Earth to Orbit: The Long Arc of Infrastructure Rivalry
The Cowboy Project occupies one point on a much longer and stranger arc of infrastructure competition between the Musk and Zuckerberg empires. In 2016, SpaceX’s rocket destroyed Facebook’s satellite — an infrastructure accident that became a personal rupture. A decade later, Tesla’s batteries store Meta’s solar power in the Wyoming high desert. The trajectory is circular in its irony and linear in its logic.
The longer view suggests that their infrastructure entanglement may extend far beyond terrestrial energy storage. Musk’s SpaceX operates Starlink, the world’s largest satellite internet constellation, which has become central to both commercial and governmental AI deployment in conflict zones, developing nations, and remote environments alike. Meta’s successive aerial and satellite connectivity programs — beginning with the Aquila solar drone initiative announced in response to the 2016 satellite loss — reflect a sustained ambition to control the connectivity layer beneath which AI services operate.
The emerging field of orbital energy infrastructure — the theoretical and increasingly practical possibility of deploying solar power generation in space and transmitting it wirelessly to ground stations — may ultimately bring these two empires into contact again, whether as competitors, partners, or both simultaneously. The rivalry that began with a rocket destroying a satellite may evolve into cooperation across a domain that does not yet have a commonly agreed name.
3.5 Strategic Interpretation: The Race for Electrons
The central strategic insight of the Meta-Tesla arrangement can be stated with compression but demands to be read at full length:
— The Core Thesis of This Paper:
The AI race is not just about smarter models. It is about who can secure electrons fastest — and who is willing to secure them from anyone, including their most visible enemy.
The IEA found that the pipeline of conditional offtake agreements between data center operators and small modular nuclear reactor projects grew from 25 gigawatts at the end of 2024 to 45 gigawatts by April 2026 — an increase of 80 percent in approximately sixteen months.27 The Brookings Institution documented that in the U.S., hyperscalers including Google, Meta, and Amazon were estimated to spend $364 billion on data center construction in 2025 alone — and that Big Tech companies accounted for 43 percent of all clean energy power purchase agreements signed globally in 2024.28
In this environment, the Cowboy Project’s structure is not merely a pragmatic energy deal. It is a template — the shape of how AI infrastructure will be powered, at scale, across the next decade. And it happens to involve two companies whose chief executives cannot stand each other.
Section 4: The Sleeping with the Frenemy Framework
The Meta-Tesla arrangement, placed within the historical context established across the preceding sections, is best understood not as an anomaly but as the most dramatic current expression of a structural pattern that demands its own analytical framework. The Sleeping with the Frenemy model identifies six conditions under which competitive rivalry yields to pragmatic cooperation, even between actors who maintain active public hostility toward one another. These six pillars are not sequential — they operate simultaneously, and their force compounds when multiple conditions are present at once.
What gives this framework its analytical power is precisely the absence of sentiment from its logic. None of its six pillars requires that rivals like each other, trust each other, or intend to remain partners permanently. Professors Qiang Ma, Hong Chen, and colleagues at Wiley’s Systems Research and Behavioral Science (2025) confirmed in their multi-case study of AI and data-driven coopetition that digital platform companies can “strategically combine cooperation and competition when leveraging AI technologies and big data resources,” finding that the unique attributes of AI resources influence coopetitive strategy in ways that challenge traditional resource-based theories.29
Pillar 1: Shared Destiny — The Infrastructure Horizon Is Identical
The first and most fundamental condition for frenemy capitalism is the recognition by both parties that their futures are, in relevant domains, structurally intertwined — not because they share values, but because they inhabit the same physical constraints. In the AI economy, the primary resources of competitive advantage are computation, energy, and connectivity. No single company, regardless of its resources, can fully vertically integrate across all three at the required scale and speed. Companies that attempt maximal self-sufficiency in all three simultaneously will be slower and more capital-inefficient than those that identify, pragmatically, which resources they must acquire externally — even from adversaries.
The World Bank’s Digital Progress and Trends Report 2025 articulated this shared destiny in geopolitical terms: high-income countries host 77 percent of global co-location data center capacity, while low-income countries account for less than 0.1 percent.30 The infrastructure arms race is not merely between competing companies — it is reshaping the global distribution of AI capability, power, and ultimately sovereignty. The companies that understand their shared destiny at a civilizational scale are best positioned to navigate the coopetitive landscape.
Pillar 2: Infrastructure Dependency — One Rival Controls What the Other Cannot Efficiently Build
The second pillar is the most concrete: one rival controls or produces something the other cannot efficiently replicate internally within the required timeframe. Infrastructure dependency — whether in chips, cloud capacity, launch systems, battery storage, or compute clusters — creates a structural opening for cooperation that transcends personal or ideological preference.
Tesla’s Megapack infrastructure represents exactly this dynamic. Meta could theoretically develop proprietary battery storage technology. But the timeline, capital requirement, and opportunity cost of doing so — relative to contracting with the established market leader — make internal development irrational. Tesla has years of manufacturing scale, supply chain integration, safety data, and operational learning embedded in its Megapack product line that no new entrant could replicate in the timeframes that matter for AI competition.
The OECD, in its 2025 analysis of competition in AI infrastructure, noted that these dependencies extend from hardware through software: “The use of open-source licences has increasingly been used as a strategic tool by companies seeking to overcome the ecosystem advantages that a technology first mover has gained,” reflecting how even the most sophisticated actors must navigate a landscape in which rivals control critical infrastructure layers.31
Pillar 3: Speed Over Pride — The Cost of Delay Exceeds the Cost of Cooperation
The third pillar is perhaps the most psychologically significant: in the AI economy, the cost of delay has become so high that it systematically overrides the emotional resistance to cooperating with rivals. Pride — the preference for competitive purity over operational efficiency — is a luxury that only companies with ample time can afford. In the AI race of 2026, time is not a luxury anyone has.
Meta’s Q1 2026 earnings reveal the stakes with devastating clarity. The company raised its full-year capex guidance in a single quarter by $10 billion — from a midpoint of $125 billion to $145 billion — because the cost of not moving fast enough on AI infrastructure was deemed greater than the cost of accelerating capital deployment.32 When an analyst on the Q1 2026 earnings call asked Zuckerberg about signs of return on the AI investment, the CEO replied:
— Mark Zuckerberg, Meta Q1 2026 Earnings Call (April 29, 2026), on ROI from AI spending:
“That’s a very technical question.”
The response — simultaneously evasive and revealing — underscored a CEO who has decided that speed of infrastructure deployment, not near-term ROI visibility, is the primary competitive metric. The Wyoming batteries that Meta will fill with Tesla Megapacks are not a philosophical choice. They are a time optimization.
Pillar 4: Capital Efficiency — Renting Can Outperform Ownership
The fourth pillar follows directly from the third: when the cost of building something internally exceeds the cost of procuring it externally — even from a competitor — procurement is the rational choice. The economic principle is straightforward; its application to corporate rivalry is psychologically uncomfortable but analytically inescapable.
The Apple-Google Gemini deal illustrates the same logic at a different layer of the AI stack. Google had invested tens of billions of dollars developing large language model infrastructure that would have taken Apple years and comparable capital to replicate independently. As Harvard Business School’s Karim Lakhani has argued, in the AI-driven firm:
— Professor Karim R. Lakhani, Harvard Business School, Digital Data Design Institute (2020):
“In this new AI-driven operating model, employees do not deliver the product or service anymore. They design and oversee the algorithms and software that do.”
The capital efficiency pillar explains why frenemy cooperation is not limited to energy infrastructure. When the intellectual infrastructure of a competitor — its AI models, its battery technology, its launch capacity, its cloud compute — is more efficiently procured than built, procurement wins. And when procurement means buying from a rival, the framework demands it.
Pillar 5: Strategic Optionality — Collaborate Now, Compete Later
The fifth pillar addresses the temporal dimension of frenemy cooperation: these arrangements are almost always explicitly or implicitly temporary. The companies involved are not surrendering their competitive ambitions. They are purchasing time — time to build their own capabilities, time to reach scale, time to move into the next competitive phase. Frenemy cooperation preserves future flexibility precisely because it does not require either party to abandon their longer-term competitive strategy.
Professor Rita McGrath of Columbia Business School has argued persuasively that in the AI era, the objective of strategy must shift from sustaining competitive positions to developing the organizational capacity to “find, exploit, and exit advantages on the organization’s own terms.”33 Strategic optionality is exactly this: the capacity to cooperate now without foreclosing the option to compete later. The company that procures from a rival while simultaneously investing in alternative supply chains is not vulnerable — it is sophisticated. The company that becomes permanently reliant on a single rival supplier has crossed from strategic cooperation into dangerous dependency.
Pillar 6: Survival Rationality — When Cooperation Is Existential
The sixth and final pillar is the most extreme: there are conditions under which refusing to cooperate with a rival is not merely inefficient — it is existentially dangerous. When the resource controlled by a rival is sufficiently critical, and when no alternative source exists at the required scale, cooperation becomes necessary for survival.
This was the logic of the 1997 Apple-Microsoft deal: Apple was not merely inefficient without Microsoft’s investment — it was ninety days from collapse. A similar logic, at a less extreme pitch, operates in the AI energy market of 2026. Companies that cannot secure reliable power for their AI data centers will not merely grow more slowly — they will be unable to train and run the AI systems that are increasingly central to their core business functions.
The IMF’s researchers Bogmans, Ganpurev, and colleagues found that under infrastructure-constrained scenarios, AI-driven electricity demand could increase U.S. electricity prices by 8.6 percent — a figure that, at the scale of Meta’s projected $125–145 billion in 2026 capex, translates into billions of dollars of additional cost.34 Survival rationality does not require that the threat be immediate. It requires only that the trajectory, if uncorrected, leads to a position from which recovery is impossible or prohibitively costly. Meta’s recognition that its AI infrastructure needs require external partnership — regardless of who the partner is — reflects exactly this rational anticipation of existential constraint.
Conclusion: The AI Economy Rewards Pragmatists, Not Purists
Meta and Tesla are not friends. Mark Zuckerberg and Elon Musk are not ideological allies. They have spent the better part of a decade as each other’s most theatrical adversaries — trading philosophical insults over the future of artificial intelligence, performing corporate hostility through the deletion of social media pages and the staging of a cage match challenge that consumed global attention without producing a single punch. They have constructed empires that are, in many domains, genuinely competitive with one another across social media, AI models, autonomous systems, and energy infrastructure.
And yet, in the Wyoming high desert near Cheyenne, a $1.2 billion energy project is taking shape in which Tesla batteries will store Meta’s solar power and release it, night after night, to keep Meta’s AI data centers running. The man who once declared that the man buying those batteries had a “limited” understanding of artificial intelligence is now, through his company, supplying the physical infrastructure without which that buyer’s AI ambitions cannot be realized. The irony has a grandeur that is almost literary.
What this paper has argued is that the Meta-Tesla arrangement is not an anomaly. It is the most vivid current expression of a structural pattern that the Sleeping with the Frenemy framework captures: infrastructure economics systematically override the emotional resistance to cooperation between adversaries when six conditions are simultaneously present — Shared Destiny, Infrastructure Dependency, Speed Over Pride, Capital Efficiency, Strategic Optionality, and Survival Rationality.
The IEA’s April 2026 findings are unambiguous about the trajectory: the capital expenditure of five major technology companies exceeded $400 billion in 2025 and is set to increase by a further 75 percent in 2026. AI-focused data center electricity consumption grew 50 percent in 2025 alone. The pipeline of nuclear energy offtake agreements for data centers grew 80 percent in sixteen months. These are not marginal trends — they are civilizational forces.35
The UNDP’s Next Great Divergence report reminds us that these forces are not distributing their benefits equitably. The same energy infrastructure arms race that is driving Meta to procure Tesla batteries for Wyoming is simultaneously concentrating computing power, AI capability, and economic advantage in a handful of wealthy nations and a handful of corporations within them. As the UNDP concluded:
— UNDP, The Next Great Divergence: Why AI May Widen Inequality Between Countries (December 2025):
“AI is becoming the general-purpose infrastructure of the 21st century, as fundamental as electricity or roads. It is critical that we don’t allow access to this infrastructure to be deeply unequal.”
The World Bank’s Digital Progress and Trends Report 2025 warns that progress is “uneven, with gaps in access and capacity that demand urgent attention” — and that without deliberate policy intervention, the AI revolution risks reversing decades of development convergence between nations.36
Against this backdrop, consider what tomorrow may hold. Today, Meta buys Tesla batteries. Tomorrow, AI rivals may rent each other’s data centers, license each other’s AI models, share orbital energy infrastructure, co-develop autonomous logistics networks, and jointly operate manufacturing facilities for the robots that will build the next generation of AI hardware. The boundaries between competition and cooperation in the AI economy are not merely blurring — they are dissolving, replaced by a more complex topology in which the same two companies can be adversaries in one domain and indispensable partners in another, simultaneously and permanently.
Tesla’s Elon Musk captured the structural inevitability of this moment on his own Q1 2026 earnings call — perhaps without fully registering that one of his most prominent customers, and one of his most famous personal rivals, was about to become precisely that:
— Elon Musk, Tesla Q1 2026 Earnings Call (April 22, 2026):
“We are making the necessary investments that will ensure our access to key materials and componentry in each region across vehicle, energy and AI as trade and geopolitics become more uncertain.”
The AI economy does not reward the company with the most principled stance toward competition. It rewards the company with the most electrons. It rewards the company with the most compute. It rewards the company that can move fastest from ambition to infrastructure to operational capability — and that is willing to do whatever it takes, including collaborating with its most visible enemy, to close the gap between what it aspires to build and what its current infrastructure permits it to build.
Sleeping with the Frenemy is not an anomaly.
— The Conclusion of This Framework:
It is the default operating model of the AI economy. The companies that understand this — and act accordingly — will define the next decade. The companies that do not will discover, too late, that in the race for artificial superintelligence, pride is a luxury that the power grid cannot afford.
Endnotes & Bibliography:
1. Bogmans, C., Ganpurev, G., Gomez-Gonzalez, P., Melina, G., Pescatori, A., & Thube, S. (April 22, 2025). Power Hungry: How AI Will Drive Energy Demand. IMF Working Paper WP/25/81. International Monetary Fund. https://www.imf.org/en/Publications/WP/Issues/2025/04/21/Power-Hungry-How-AI-Will-Drive-Energy-Demand-566304
2. Enbridge Inc. (May 2026). Enbridge and Meta Expand Partnership with Cowboy Project. Official Press Release. https://www.enbridge.com/stories/2026/may/enbridge-meta-expand-partnership-cowboy-project-solar-plus-battery-storage
3. Meta Platforms, Inc. (April 29, 2026). Meta Q1 2026 Earnings Call Transcript. The Motley Fool. https://www.fool.com/earnings/call-transcripts/2026/04/29/meta-meta-q1-2026-earnings-call-transcript/
4. Cult of Mac. (August 6, 2025). Today in Apple History: Microsoft Throws Apple a $150 Million Lifeline. https://www.cultofmac.com/news/microsoft-investment-saves-apple
5. Strategy Breakdowns. (October 2025). How Microsoft Saved Apple. https://strategybreakdowns.com/p/how-microsoft-saved-apple
6. CNBC. (January 12, 2026). Apple Picks Google’s Gemini to Run AI-Powered Siri Coming This Year. https://www.cnbc.com/2026/01/12/apple-google-ai-siri-gemini.html
7. Marketing Dive. (January 15, 2026). Apple Taps Google Gemini to Power AI Features in Multiyear Deal. https://www.marketingdive.com/news/apple-taps-google-gemini-to-power-ai-features-in-multiyear-deal/809697/
8. OECD. (November 2025). Competition in Artificial Intelligence Infrastructure. OECD Roundtables on Competition Policy Papers. https://www.oecd.org/en/publications/2025/11/competition-in-artificial-intelligence-infrastructure_69319aee.html
9. World Bank. (November 2025). Digital Progress and Trends Report 2025: Strengthening AI Foundations. https://www.worldbank.org/en/publication/dptr2025-ai-foundations
10. NBC News. (September 1, 2016). SpaceX Rocket Explosion Destroys Facebook Internet Satellite. https://www.nbcnews.com/tech/tech-news/spacex-rocket-explodes-launch-pad-n641261
11. Fortune. (July 25, 2017). Elon Musk Just Dissed Mark Zuckerberg’s Understanding of Artificial Intelligence. https://fortune.com/2017/07/25/elon-musk-just-dissed-mark-zuckerbergs-understanding-of-artificial-intelligence
12. Fortune. (March 31, 2026). 2 Years After Musk Challenged Zuckerberg to a Cage Match, They Were Texting About DOGE and a Joint OpenAI Bid. https://fortune.com/2026/03/31/elon-musk-mark-zuckerberg-doge-openai-takeover-court-documents/
13. Psychology Today. Travers, M. Ph.D. (September 13, 2023). Why the Musk-Zuckerberg Duel Is a Snapshot of Our Culture. https://www.psychologytoday.com/us/blog/social-instincts/202309/why-the-musk-zuckerberg-duel-is-a-snapshot-of-our-culture
14. Ibid. Travers, M. Ph.D. (2023). Psychology Today. https://www.psychologytoday.com/us/blog/social-instincts/202309/why-the-musk-zuckerberg-duel-is-a-snapshot-of-our-culture
15. Meta Platforms, Inc. (April 29, 2026). Meta Reports First Quarter 2026 Results. Official Investor Relations. https://investor.atmeta.com/investor-news/press-release-details/2026/Meta-Reports-First-Quarter-2026-Results/default.aspx
16. Tesla, Inc. Form 10-Q Q1 2026. Filed with the U.S. Securities and Exchange Commission. https://www.sec.gov/Archives/edgar/data/0001318605/000162828026026673/tsla-20260331.htm
17. Yahoo Finance / CoinDCX. (April 22, 2026). Tesla Q1 2026 Earnings Results: EPS Beats on Margin Rebound. https://coindcx.com/blog/us-stock/tesla-q1-2026-earnings-results/
18. Utility Dive. (May 2026). Enbridge, Meta to Build 365-MW Solar/200-MW Storage Project. https://www.utilitydive.com/news/enbridge-meta-build-365-mw200-mw-solarstorage-project/820905/
19. BIC Magazine. (May 2026). Enbridge Cowboy Project: 365MW Solar & 1600MWh Battery for Meta in Wyoming. https://www.bicmagazine.com/projects-expansions/renewable-sustainability-h2-esg/enbridge-meta-cowboy-project/
20. Carbon Credits. (May 2026). Meta’s AI Power Surge Drives $1.2 Billion Solar Megaproject with Enbridge. https://carboncredits.com/metas-ai-power-surge-drives-1-2-billion-solar-megaproject-with-enbridge/
21. Energy-Storage.news. (October 2025). Tesla Reports Record Energy Storage Deployments. Megapack 3/Megablock launch at RE+ 2025. https://www.energy-storage.news/tesla-reports-record-energy-storage-deployments-and-profit-ahead-of-vote-on-musks-monster-pay-proposal/
22. Tesla Accessories/TESMAG. (2026). Tesla Energy Q1 2026 Update: Megapack Deployments Surge to Record Highs. https://www.teslaacessories.com/blogs/news/tesla-energy-q1-2026-update-megapack-deployments-surge-to-record-highs-as-utility-scale-storage-transforms-the-grid
23. International Energy Agency. (April 16, 2026). Data Centre Electricity Use Surged in 2025. https://www.iea.org/news/data-centre-electricity-use-surged-in-2025-even-with-tightening-bottlenecks-driving-a-scramble-for-solutions
24. IEA. (April 2026). Key Questions on Energy and AI. Executive Summary. https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary
25. Brookings Institution. (April 2, 2026). Global Energy Demands Within the AI Regulatory Landscape. https://www.brookings.edu/articles/global-energy-demands-within-the-ai-regulatory-landscape/
26. UNDP. (December 2, 2025). The Next Great Divergence: Why AI May Widen Inequality Between Countries. https://www.undp.org/asia-pacific/next-great-divergence
27. IEA. Key Questions on Energy and AI. (April 2026). SMR pipeline growth: 25 GW to 45 GW in 16 months. https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary
28. Brookings Institution. (April 2026). Global Energy Demands: Big Tech accounts for 43% of all global clean energy PPAs in 2024. https://www.brookings.edu/articles/global-energy-demands-within-the-ai-regulatory-landscape/
29. Ma, Q., Chen, H., Tian, S., et al. (2025). Navigating Coopetition: A Multiple Case Study of AI and Data-Driven Strategies in the Digital Platform Economy. Systems Research and Behavioral Science, Vol. 42, Issue 4, pp. 1043–1060. Wiley. https://onlinelibrary.wiley.com/doi/abs/10.1002/sres.3067
30. World Bank. (November 2025). Digital Progress and Trends Report 2025. High-income countries host 77% of global co-location data center capacity. https://www.worldbank.org/en/publication/dptr2025-ai-foundations
31. OECD. (November 2025). Competition in Artificial Intelligence Infrastructure. Open-source licensing as strategic tool. https://www.oecd.org/en/publications/competition-in-artificial-intelligence-infrastructure_623d1874-en/full-report/component-6.html
32. Fortune. (April 29, 2026). Meta Just Bumped Its 2026 Capex Forecast Up to as Much as $145 Billion. https://fortune.com/2026/04/29/meta-zuckerberg-145-billion-ai-spending-roi/
33. McGrath, R. G. (2013). The End of Competitive Advantage. Harvard Business Review Press. See also: HBR Strategy Summit 2026, Columbia Business School (February 2026). https://business.columbia.edu/staff/people/rita-mcgrath
34. Bogmans et al. IMF WP/25/81. (April 2025). U.S. electricity prices could increase by 8.6% under constrained infrastructure scenarios. https://www.imf.org/en/Publications/WP/Issues/2025/04/21/Power-Hungry-How-AI-Will-Drive-Energy-Demand-566304
35. IEA. (April 2026). Key Questions on Energy and AI. $400B capex by five tech firms in 2025; 75% further increase projected for 2026. https://www.iea.org/news/data-centre-electricity-use-surged-in-2025-even-with-tightening-bottlenecks-driving-a-scramble-for-solutions
36. World Bank. (November 2025). Strengthening AI Foundations: Emerging Opportunities for Developing Countries. https://www.worldbank.org/en/news/factsheet/2025/11/21/strengthening-ai-foundations-emerging-opportunities-for-developing-countries
