AI CapEx, Circular Revenues, Stablecoins—and the Career Math Behind Big Bets
AI CapEx, Circular Revenues, Stablecoins—and the Career Math Behind Big Bets
Author: Zion Zhao Real Estate | 88844623 | 狮家社小赵
Author’s note: This essay builds on my fundamental study for the day and elaborate on the implications for technology, finance, policy, and careers. I write and revise pieces like this daily—hours of primary reading, interviews, podcasts, filings, and data—because due diligence matters.
1) Are we in an AI bubble—or financing the next computing era?
The headline numbers are staggering. Depending on the forecaster, cumulative AI-related data-center capital expenditure (capex) over the second half of the 2020s ranges from ~US$1.2T to ~US$3T–4T. Dell’Oro Group now projects global data-center capex to reach about US$1.2T by 2029 (with hyperscalers ~50%) (Dell’Oro Group, 2025). Citi, in contrast, has lifted its cumulative forecast to US$2.8T through 2029, and several executives—including NVIDIA’s Jensen Huang—have discussed industry scenarios in the US$3T–4T by ~2030 range (Citi via The Australian; Reuters, 2024–2025). The dispersion tells you two things: confidence in demand is high, but the slope and timing are uncertain. Fierce Network+2The Australian+2
Under the hood, at least three drivers justify heavy front-loaded capex even if you’re skeptical of “bubble” narratives:
Compute shifts from CPU to accelerated computing for AI training and inference. Multiple years of replacement cycles are likely as accelerated compute becomes the default for more workloads (Huang in Reuters, 2024–2025). ai-supremacy.com
Energy and memory are now first-class bottlenecks. Power availability and HBM memory capacity/packaging constrain deployments; securing these inputs pulls spend forward.
Jevons-like effects—as tokens/TFLOPs per dollar fall, new use-cases become economical, expanding demand (a pattern seen with bandwidth in the 2000s).
That said, the how of financing today’s buildout matters as much as the how much. Which brings us to revenue quality and “round-tripping.”
2) “Circular” revenues, credits, and off-take guarantees: what is healthy vs. hype?
Investors and auditors flag a spectrum:
On one end: Sham round-trip deals—no economic substance, simply cycling cash/products to book revenue—are classic red flags in SEC enforcement history (see early-2000s cases) (U.S. SEC, 2003). The Wall Street Journal
In the middle: Induced demand—equity or credits that materially condition a purchase/usage decision—may still be lawful but lowers revenue quality. Recent examples drawing scrutiny include cloud credits tied to strategic investments (e.g., AWS–Anthropic; regulators and press have noted the accounting/competition questions) (Reuters, 2023). GeekWire
On the other end: Strategic investments in well-capitalized buyers with many funding options (where products would likely be purchased regardless) look far healthier.
Two current flashpoints:
Cloud credits at scale. Microsoft, Amazon, and Google have extended investment-linked credits to AI model companies to drive platform usage. The practice isn’t per se improper; it’s the materiality test that matters—would usage at this level have happened “but for” the credits? (Reuters, 2023). GeekWire
Off-take/guarantee structures in the neo-cloud layer. A recently surfaced provision indicates NVIDIA will guarantee compute capacity availability for CoreWeave through 2032, a commitment that can improve CoreWeave’s financing terms but may muddy outsiders’ ability to read “pure” demand signals if unused capacity can be redirected (Yahoo Finance summary, 2025). Here, disclosure and analyst diligence are critical. delloro.com
Practical test for investors:
Economic substance: Is there downstream, third-party end demand?
Transfer pricing & arm’s-length: Would the counterparty buy at this price absent the investment/credit?
Disclosure hygiene: Are obligations (credits, minimums, off-take, step-ins) clearly disclosed?
Balance-sheet resilience: If demand softens, who’s left holding fixed commitments?
History suggests that overbuilds usually surface first at the periphery (levered, single-customer-dependent nodes), not among the hyperscalers. Healthy skepticism is warranted—but so is nuance.
3) Regulation: the case for federal preemption (and measured guardrails)
Two state-level moves crystallize the risks of a 50-state patchwork for interstate technologies:
Colorado’s AI Act (SB24-205) creates obligations around “algorithmic discrimination” for “high-risk” systems; it is broadly framed and shifts duties up the supply chain—including providers of general-purpose models (Colorado General Assembly, 2024). Seyfarth Shaw - Homepage
California’s SB 243 (2025) imposes safety, security, content, and record-keeping duties on “chatbot companions,” with a private right of action (civil suit) for harms—raising litigation and compliance uncertainty, especially for startups (California Legislative Info, 2025). LegiScan
Meanwhile, Governor Newsom vetoed SB 1047 (2024)—a sweeping frontier-model bill—citing innovation concerns and the EU AI Act’s extraterritorial reach as context for preferring targeted, risk-based approaches (Office of Governor Newsom, 2024).
A sensible path forward is federal preemption plus risk-proportional guardrails, anchored in frameworks like NIST’s AI Risk Management Framework (AI RMF 1.0)—voluntary but increasingly used as a baseline for governance, testing, and monitoring (NIST, 2023). Preemption avoids Balkanized rule-sets; RMF-style guidance reduces ambiguity while preserving velocity. static.carahsoft.com
4) Stablecoins and the new money rails: from Brazil’s PIX to FedNow to USDC “rewards”
While AI captures headlines, digital money rails may be the stealth revolution. Three developments stand out:
U.S. stablecoin market structure is maturing. The combined market cap (led by USDT and USDC) has surged past US$300–350B; importantly, fully reserved structures hold a large share of short-duration Treasuries, altering demand dynamics at the margin (Reuters, 2025). Coinbase now pays ~4.1% in USDC rewards on eligible balances—economically akin to interest from a consumer’s perspective, even if labeled “rewards” (Coinbase, 2025).
International pressure for instant, low-cost payments is real. Brazil’s PIX reached near-universal adoption within five years and is widely credited with boosting competition and lowering payment costs. The U.S. Trade Representative’s 2025 review of PIX (amid concerns from U.S. incumbents) underscores how geopolitics and competition policy intersect with consumer payments (Reuters, 2025).
The U.S. has FedNow—but network effects lag. The Federal Reserve launched FedNow in 2023 to enable instant clearing/settlement between participating banks. Coverage is expanding, but ubiquity and developer UX trail leading “instant-pay” regimes abroad, and card networks still dominate point-of-sale (Federal Reserve, 2023).
Two caveats are essential. First, throughput and resilience: VisaNet reports theoretical peaks above 65,000 TPS, whereas L1 blockchains like Ethereum remain far lower (with L2s and Solana narrowing the gap), so mainstream retail payments at full Visa scale still demand either L2s, off-chain netting, or hybrid models (CoinDesk, 2023). Second, prudential and consumer risks: the BIS has repeatedly warned that stablecoins can create “run” dynamics, settlement risk, and opacity if not properly supervised and segregated (BIS, 2024). Balance-sheet transparency and bankruptcy-remote structures remain non-negotiable. delloro.com
Bottom line: open, programmable money rails are arriving. Whether delivered by stablecoin issuers, banks using FedNow, or big-tech wallets on top of either, the practical test is simple: safer custody, clear disclosures, instant settlement, lower fees, and wide merchant acceptance.
5) Careers and capex: “Run down a dream,” but use a decision framework
One striking resonance from my discussion with my peers is how career strategy mirrors capex allocation: place bold bets, but sequence them thoughtfully. The “regret-minimization” test popularized by Jeff Bezos is an accessible heuristic: project yourself forward; which choice—taking the shot or sitting it out—minimizes lifetime regret? Research on career satisfaction corroborates that inaction regrets loom larger than action regrets over time (Pink, 2022). Meanwhile, Gallup continues to find that a minority of workers report being “engaged” at work—an indictment of grind-first, passion-later tracks that never get revisited.
Practical takeaways for builders and investors:
Stack optionality early. Skills that compound (statistics, coding, product, selling, writing) raise the floor on future pivots.
Interrogate unit economics—of your company and your career. Credits and guarantees can mask softness; so can a prestigious job with stagnant learning.
Align with durable tailwinds. AI, energy infrastructure, and programmable money are multi-cycle themes; enter through honest edges (compliance strength, distribution, or cost of capital).
Beware policy friction. State patchworks add drag—choose regulatory-aware go-to-market paths.
6) What to watch next (objective scorecard)
Capex-to-cash-flow ratios at the “MAG” platforms through 2026: do they normalize as revenue from AI products appears? (Consensus currently expects moderation after a 2025–26 peak.)
Disclosure clarity on credits, minimums, and off-take arrangements—particularly at neo-clouds and custom-silicon efforts.
Federal action on AI (targeted preemption; adoption of NIST-aligned risk frameworks) versus state-by-state accretion.
Stablecoin policy (reserve segregation, disclosures, and payment-system access) and FedNow adoption metrics.
Power and HBM supply timelines: the true governors of AI’s slope.
If the next computing platform is truly here, its hallmark will be less exuberance and more execution: clear ROI, transparent accounting, prudent regulation, and resilient infrastructure.
Build a Future-Ready Portfolio in Singapore—With a Real-Assets Advisor Who Also Speaks AI, Macro & Law
In an AI-fueled, high-beta market, balance sheet resilience starts with smart real estate.
I’m a Singapore-based real estate agent who also lives and breathes economics, global affairs, asset allocation, and portfolio construction. As a seasoned equities & crypto trader (macro + technicals) and a practitioner of Singapore Land & Business Law, I bring a multi-asset, risk-managed lens to your property decisions—backed by the discipline and integrity of an Officer Commanding (Captain), Singapore Armed Forces (SAF).
Every day I dedicate hours to researching macro trends, regulation, AI capex cycles, “circular” revenues, digital-asset rails (stablecoins), and global geopolitics—and I turn that diligence into clear, actionable property strategy for you. I write extensively so you can see my process and my homework.
Why work with a real-estate advisor who understands more than real estate?
Because your property is part of a portfolio, not an island. You deserve advice that integrates:
Macro & market timing — Entry/exit discipline aligned with rates, liquidity, and capital flows—not guesswork.
Cross-asset trade-offs — How Singapore property stacks up versus equities/AI plays/crypto on risk, yield, and drawdowns.
Risk budgeting — Stress tests on rent, vacancy, rates, FX, and policy; structuring for downside protection.
Legal & regulatory clarity — ABSD/SSD/TDSR/LTV, title & caveats, land use, tenancy, compliance—handled precisely.
Who I serve
International, Mainland China (中国大陆), Hong Kong/Macau/Taiwan, Southeast Asia, and Singapore clients—UHNW families, family offices（家办）, institutions, and parents planning education paths（陪读家长 / 留学置业）who want Singapore exposure across:
Prime residential (CCR/RCR/OCR), landed & GCBs, heritage shophouses
Yield assets: Grade-A/strata office, retail, business parks, logistics/industrial
Purpose-built student accommodation & co-living (school catchments)
Developer blocks / en-bloc potential (where appropriate and compliant)
What you can expect in our first strategy session (complimentary)
Portfolio fit: Where real estate belongs in your asset mix for lower volatility, inflation hedge, and dividend-like rental income.
Acquisition plan: Target sub-markets, school proximity, amenities, transport, tenant demand, and rental comps.
Numbers that matter: Cap rate, yield-on-cost, cash-on-cash, IRR scenarios; sensitivity to rates and policy.
Execution roadmap: Shortlist, due diligence, legal checks, financing, tax/structuring (with licensed partners), and post-purchase asset management.
Why Singapore, why now
A trusted, rules-based, AAA-rated hub with deep capital markets and strong property rights.
Tight supply in key segments, resilient tenant demand, and historically attractive risk-adjusted returns.
Portfolio benefits: lower correlation to high-volatility assets, stable rental yields, and compounding capital appreciation over a full cycle.
Ready to future-proof your wealth?
Whether you’re allocating from tech/AI gains, rebalancing out of volatile assets, or building a family base in a stable jurisdiction, let’s put a real-assets plan behind your ambitions—quant-rigor, legal precision, and military-grade execution.
Message me to schedule a confidential 30-minute strategy call.
I’ll share a concise market brief, a tailored shortlist, and a clear next-step plan—so you can act with conviction.
给国际/中国/东南亚/新加坡客户：
需要在新加坡投资自住/收租、子女就学（陪读家长/留学）、或家族办公室（家办）配置？我结合宏观经济、全球政经、股债加密多资产配置与新加坡法律框架，为您制定稳健的地产方案：波动更低、现金流可见、资本增值可期。欢迎私信预约一对一咨询。
References (APA)
Bank for International Settlements. (2024). Stablecoins: risks, potential and regulation. Basel, Switzerland: BIS. Retrieved from
California Legislative Information. (2025). SB-243 Privacy and safety for AI companion products (2025–2026). Retrieved from LegiScan
Coinbase. (2025). USDC rewards terms and rate (4.10% variable APY). Retrieved from
Colorado General Assembly. (2024). SB24-205—Artificial Intelligence and consumer protections. Retrieved from Seyfarth Shaw - Homepage
CoinDesk. (2023, Aug. 31). Visa says its network can handle over 65,000 TPS; how does that compare to blockchains?Retrieved from delloro.com
Dell’Oro Group (via Fierce Network). (2025, Aug. 6). Data center capex projected to reach $1.2 trillion by 2029. Retrieved from Fierce Network
National Institute of Standards and Technology (NIST). (2023). AI Risk Management Framework (AI RMF 1.0). Gaithersburg, MD: U.S. Department of Commerce. Retrieved from static.carahsoft.com
Office of Governor Gavin Newsom. (2024). Governor Newsom veto message on SB 1047; statement on AI policy direction and EU AI Act context. Retrieved from
Pink, D. H. (2022). The power of regret: How looking backward moves us forward. New York, NY: Riverhead Books.
Reuters. (2023, Nov. 3). U.S. FTC and regulators eye big tech AI cloud investments tied to credits. Retrieved from GeekWire
Reuters. (2024–2025). Global AI infrastructure spending could reach $3–4 trillion by 2030, industry leaders say. Retrieved from ai-supremacy.com
Reuters. (2025, Oct.). USTR examines Brazil’s PIX after U.S. firms raise concerns; stablecoin implications. Retrieved from
The Australian / Citi Research (summary). (2025, Sept.). Citi lifts cumulative AI infrastructure forecast to $2.8T through 2029. Retrieved from The Australian
U.S. Securities and Exchange Commission. (2003). The Sarbanes-Oxley Act of 2002: The first year—“Round-trip” transactions and revenue recognition in enforcement. Washington, DC: SEC. Retrieved from The Wall Street Journal
Visa Inc. (2024–2025). Visa Crypto: Stablecoin dashboard and research hub. Retrieved from
Yahoo Finance (summary). (2025, Aug.). NVIDIA to guarantee CoreWeave capacity availability through 2032. Retrieved from delloro.com
Federal Reserve. (2023). FedNow® Service—Instant payments for financial institutions. Retrieved from
Politico / The Guardian (policy coverage). (2025). Invest America (“Trump accounts”) implementation timeline and early guidance. Retrieved from
Notes on factual uncertainties (transparency)
OpenAI–Broadcom custom chip: Multiple outlets have reported discussions/partnership explorations. As of this writing, details vary by report; any specific performance or volume claims should be treated as provisional until filings or press releases provide technical and commercial specifics.
Stablecoin “$18T settled” claim: Aggregators often cite rolling 12-month on-chain transfer values. Methodologies differ (gross vs. adjusted). Prefer dashboards with transparent de-spamming/velocity adjustments (e.g., Visa’s research hub) and BIS cautionary analyses for a balanced view.
Comments
Post a Comment