To The Moon

Here's a comprehensive, bullish counter-article based on the title "Hyperscaler Earnings 'Fraud'? Reality Check: AI Investments Are Paying Off Now". It directly tackles Michael Burry's accusation of "earnings fraud" via extended depreciation on AI GPUs/servers (estimating ~$176 billion understated depreciation from 2026–2028, potentially overstating Meta earnings by ~21% and Oracle by ~27%), while shifting focus to tangible 2025–early 2026 evidence of accelerating cloud/AI revenue, backlog growth, monetization, and productivity signals from hyperscalers as of April 2026.Hyperscaler Earnings "Fraud"? Reality Check: AI Investments Are Paying Off NowMichael Burry has aggressively labeled hyperscaler accounting practices as one of the "most common forms of fraud in the modern era." His core claim: companies like Meta, Microsoft, Google (Alphabet), Amazon, and Oracle are artificially inflating earnings by extending depreciation lives on expensive Nvidia-powered GPUs and servers from a more "realistic" 2–3 years to 5–6+ years. This, he argues, understates expenses by an estimated $176 billion between 2026 and 2028, masking weak ROI on the massive AI capex surge and setting up an eventual earnings cliff.Burry's critique gained traction amid 2025–early 2026 volatility, aligning with his shorts on AI-related names. Yet as we stand in April 2026, a reality check reveals the opposite: hyperscaler AI investments are already delivering measurable payoffs through accelerating cloud revenue, expanding backlogs, rapid capacity monetization, and early productivity gains. Extended depreciation isn't sleight-of-hand—it's a rational reflection of real-world utilization, efficiency improvements, and the shift toward inference and agentic AI. Here's why Burry's "fraud" narrative underestimates the value creation unfolding today.The Depreciation Debate: Context Over AccusationBurry contends that short product cycles for AI chips (Nvidia's rapid architecture releases) demand aggressive depreciation. Extending lives, he says, artificially boosts reported profits—potentially overstating Meta by 21% and Oracle by 27% by 2028.Counterpoint: Useful economic life isn't solely about silicon obsolescence. Newer architectures (e.g., Blackwell) deliver massive efficiency gains, allowing older hardware to handle inference workloads effectively alongside newer systems. Hyperscalers report high utilization rates and sustained demand, not stranded assets. GPUs aren't discarded after 2–3 years; they complement evolving stacks in training/inference hybrids, sovereign projects, and enterprise deployments. This mirrors how cloud infrastructure investments in prior cycles proved more durable than initial skeptics expected.Importantly, hyperscalers maintain strong overall cash flows and operating income growth despite heavy spending. The accounting choice reflects operational reality—extended usefulness driven by software optimizations, networking advances, and compounding AI use cases—rather than manipulation.AI Investments Paying Off: Revenue Acceleration in 2025–Q4 2025/Q1 2026Far from a masked weakness, hyperscaler earnings show clear AI-driven momentum:Amazon (AWS): Q4 2025 cloud revenue hit $35.6 billion, up 24% YoY—the fastest growth in 13 quarters. AWS operating income rose to $25.0 billion. CEO Andy Jassy highlighted AI innovation driving demand, with advertising and chips businesses also contributing strongly. Full-year AWS trends pointed to sustained acceleration.

Microsoft (Azure/Intelligent Cloud): Cloud revenue grew ~26% in recent quarters, with Azure and other cloud services up 39% in one reported period. Microsoft Cloud overall showed robust expansion, fueled by Copilot adoption, enterprise AI tools, and OpenAI synergies. Backlogs remain elevated, with capacity constraints signaling demand outpacing supply.

Alphabet (Google Cloud): Q4 2025 cloud revenue reached $17.7 billion, up 48% YoY—the standout performer among the big three. Growth was driven by AI infrastructure, Vertex AI, and enterprise demand. Backlogs surged, and AI-related products (Gemini, etc.) showed triple-digit or higher gains in segments.

Meta: While more ad-focused, AI improvements in content recommendation and ad relevance supported strong operating income guidance despite capex ramps. Meta's Llama models and internal AI investments contributed to efficiency and engagement lifts.

Collectively, the big three clouds (AWS, Azure, Google Cloud) generated over $86 billion in Q4 2025 revenue, with accelerating growth rates. AI workloads are a key driver: inference is inflecting, agentic systems are emerging, and enterprises are moving from pilots to production. Nvidia's own Q4 FY2026 results (ended Jan 2026) reinforced this—data center revenue of $62.3 billion (up 75% YoY), total revenue $68.1 billion (up 73%), with Q1 FY2027 guidance at $78 billion. Hyperscalers remain the largest customers, but diversification into sovereign AI and enterprise adds resilience.Monetization and ROI Signals: Beyond AccountingBurry's focus on near-term optics overlooks forward indicators:Backlogs and visibility: Microsoft noted significant Azure backlogs due to power constraints. Amazon highlighted rapid capacity addition (e.g., gigawatts doubling). Google Cloud backlogs grew sharply.

Productivity and value creation: Early data shows 25–40% task-level gains in coding, analysis, and operations. Hyperscalers report improved ad performance, cloud service uptake, and internal efficiencies offsetting some capex costs.

Capex context: The ~$700 billion combined 2026 guidance (Amazon ~$200B, Alphabet $175–185B, Meta $115–135B, Microsoft ~$120B+) is enormous but funded by massive operating cash flows. Many continue guiding for operating income growth. Temporary FCF pressure exists for some, but this is infrastructure investment akin to electricity grids or broadband—not speculative excess.

Unlike dot-com-era hype with thin monetization, today's AI delivers enterprise value now: copilots, optimized workloads, automation, and data-driven decisions. The shift from training-heavy to inference-heavy spend in 2026 further improves economics.Risks and the Balanced ViewLegitimate concerns remain: execution on ROI timelines, energy/power bottlenecks, competition in custom silicon, or macro slowdowns could delay payoffs. Heavy 2026 capex may pressure free cash flow for names like Amazon. Valuations embed high expectations, and Burry's track record demands respect on volatility.Yet the data as of April 2026—accelerating cloud growth, Nvidia's demand visibility, and hyperscaler commentary on rapid monetization—suggests investments are paying off sooner than skeptics anticipated. Depreciation debates are accounting nuances; the underlying demand and value creation are structural.The Bottom Line: Payoffs Are Here, Not HypotheticalBurry's "fraud" warning paints a picture of illusory earnings propped up by tricks. Reality shows hyperscalers generating record cloud revenues, expanding backlogs, and positioning for an AI-native economy where compute translates into tangible business outcomes.AI investments aren't a future promise—they're delivering acceleration today. As inference, agents, and enterprise adoption scale, the ROI case strengthens. Burry may highlight valid accounting questions, but calling it fraud overlooks the breakthrough economics unfolding in data centers and boardrooms worldwide.The real risk for bears? 

$Palantir Technologies Inc.(PLTR)$  $NVIDIA(NVDA)$  $Amazon.com(AMZN)$  $Broadcom(AVGO)$