Nvidia (NVDA) vs Fed Balance Sheet

Nvidia is now the clearest single-stock proxy for the AI capex cycle: hyperscaler capex commitments from Microsoft, Amazon, Alphabet, Meta and a widening set of sovereign and tier-2 cloud buyers route through Nvidia GPUs, with multi-year lead times on Blackwell and Rubin platform deliveries. WALCL captures the Federal Reserve System's total assets less eliminations from consolidation, the headline number on the H.4.1 statistical release that the Federal Reserve Board has published every Thursday since 1996.

The macro question the spread answers is whether liquidity creation by the central bank is the binding constraint on Nvidia's multiple, or whether the AI capex flywheel has decoupled NVDA from the central-bank liquidity backdrop. From January 2022 through October 2022, WALCL contracted by roughly $300 billion as the Fed began QT, and NVDA fell about 65 percent peak-to-trough. From early 2023 through late 2024, WALCL contracted by another $1.4 trillion while NVDA rose more than 800 percent. That single regime contradicts the textbook QE-equals-multiple-expansion script and is what makes the pair worth tracking explicitly.

On December 18, 2024 the FOMC announced that quantitative tightening would slow further in early 2025, capping monthly Treasury runoff at $5 billion from the prior $25 billion and leaving the agency MBS cap at $35 billion. The Fed ended balance-sheet runoff entirely in December 2025 and now reinvests maturing Treasuries to match trend reserve demand, with MBS holdings still grinding lower as principal returns. The result is that WALCL has stabilized in a narrow band between $6.6 and $6.8 trillion through the first four months of 2026.

Nvidia's behavior since the QT halt has been driven much more by capex disclosures than by the balance sheet line. Microsoft's April 29, 2026 fiscal Q3 print guided full-year capex toward $190 billion, up 61 percent year-over-year and roughly $35 billion above the Visible Alpha consensus of $154.6 billion. Each $10 billion of incremental hyperscaler capex translates roughly to $4 to $5 billion of Nvidia data-center revenue at current allocation ratios. With WALCL flat and capex accelerating, NVDA has reached new highs while the balance-sheet leg has barely moved, which is exactly the configuration that decouples the pair.

Nvidia's modeled valuation is unusually back-loaded: sell-side DCF templates from Morgan Stanley, JP Morgan and Goldman Sachs typically place 70 to 85 percent of NVDA's equity value in cash flows beyond year five, reflecting the long monetization tail on AI inference workloads and sovereign AI build-outs. That structure makes NVDA mathematically more sensitive to changes in the 10-year real yield than to changes in WALCL itself. The 10-year TIPS yield ran near 2.0 percent through April 2026, well above the 0 to 1 percent band that prevailed during the 2020-2021 NVDA run.

The practical implication is that the WALCL leg of the pair is best read as a leading indicator of the policy reaction function rather than as a direct discount-rate input. When WALCL expansion telegraphs a coming easing cycle, real yields fall and NVDA rerates. When WALCL is flat but the FOMC pivots verbally toward easing, real yields can fall without any balance-sheet move, and NVDA can rerate without WALCL moving at all. Episodes where both legs move together remain the cleanest, but they have become rarer since the December 2025 QT halt.

The 2022 episode is the single most informative period for the pair. The FOMC began balance-sheet runoff on June 1, 2022 with caps phased in at $30 billion Treasuries and $17.5 billion MBS, then doubling on September 1 to $60 billion and $35 billion. WALCL fell from $8.965 trillion on April 13, 2022 to $8.583 trillion by year-end, a contraction of roughly $382 billion. NVDA peaked at $33.38 (split-adjusted) in November 2021, fell to $11.20 in October 2022, and the pair traded as a clean liquidity story for about ten months.

The pair broke from that template starting in late January 2023. ChatGPT's November 2022 launch had moved AI from a research curiosity to a commercial earnings driver, and Nvidia's May 24, 2023 fiscal Q1 2024 print, which guided to fiscal Q2 revenue of $11 billion against consensus of $7.2 billion, marked the moment when the AI capex story decisively overrode the liquidity narrative. WALCL kept contracting, but NVDA stopped trading on the balance sheet.

The current configuration is ambiguous in the information-content sense: WALCL is flat at $6.70 trillion, the QT regime has ended, and the AI capex story is providing the dominant signal. Practitioners using the pair are watching two specific triggers. First, any reversal back to outright balance-sheet expansion, signaled in advance through the H.4.1 weekly print and confirmed by FOMC minutes, would produce a coordinated move higher in both legs and is the historical setup that has produced the strongest NVDA outperformance.

Second, a meaningful slowdown in hyperscaler capex commitments, visible through fiscal-quarter reports from Microsoft, Amazon, Alphabet and Meta in late April and early May 2026, would expose NVDA to the discount-rate channel without the offsetting capex tailwind. The April 29 Microsoft print pointed in the opposite direction with the $190 billion capex guide, but Wall Street is already split on whether the Azure acceleration justifies the spend, and the 24/7 Wall St analyst summary on April 30 flagged the bull-bear divide explicitly. Trading the pair right now means watching capex calls more closely than watching the H.4.1.

The Convex Net Liquidity Impulse decomposes WALCL into its three economically distinct components: outright Federal Reserve assets, the Treasury General Account, and reverse repo balances. WALCL alone can be misleading because TGA and RRP movements can offset each other inside a flat balance-sheet number. During the 2023-2024 stretch when WALCL contracted by $1.4 trillion, the RRP facility drained from $2.55 trillion in late 2022 to under $200 billion by November 2024, releasing roughly $2.3 trillion of reserves into the banking system without changing WALCL.

Reading NVDA against the net liquidity impulse rather than headline WALCL produces a much tighter relationship through the 2023-2024 regime. The decomposition is the operational reason analysts who tracked CNLI rather than WALCL alone caught the AI rerate earlier and held it longer. For the April 2026 setup, the net liquidity impulse is essentially flat with TGA running near $700 billion and RRP near $150 billion, leaving the AI capex line as the dominant near-term signal.