August was expected to be a quiet month for markets, yet it has instead become a profound release of policy signals.

US Treasury Secretary Scott Bessent has intervened in both the dollar-yen exchange rate and the long end of the Treasury market, sending a clear message to investors: authorities would rather accept currency depreciation than allow the bond market to collapse. This shift has reshaped how investors position themselves, with gold, bitcoin, and industrial metals returning to the spotlight, while the artificial intelligence narrative accelerates its transition from a "construction phase" to an "application phase." Goldman Sachs partner Mark Wilson has characterized this series of events as a "consequential series of August events."

Wilson believes Bessent's intervention represents the most significant policy milestone of the month. The Treasury Secretary first stepped into the dollar-yen pair, then pushed for long-end Treasury repurchase operations — the so-called "twist" — while simultaneously expanding bank balance sheet capacity and shrinking the Federal Reserve's balance sheet, demonstrating that the policy toolbox is far from exhausted. The Goldman Sachs rates team notes that while these repo operations may improve long-end supply-demand dynamics at the margin, they do not address the fundamental driver pushing global—not just US—long-end yields higher.

The market's response to this policy signal has been swift and direct. Over the past ten days, investors have shown palpable anxiety over their under-exposure to "store of value" assets, with gold, bitcoin, and metals investment logic returning to the center of every investor conversation.

Meanwhile, NVIDIA saw its market capitalization surge by nearly $500 billion in a single day, and the software ETF IGV recorded its best single-day performance in two decades. Yet the S&P 500, excluding AI-related stocks, fell 0.7% that same day — a market breadth reading so poor it ranks among the worst this century. Two distinctly different narratives are now running in parallel within the same market.

The Policy Hand: What Bessent's "Twist" Really Means

The persistent rise in long-end yields is the most visible risk facing an already extended bull market in equities, particularly against the backdrop of deteriorating fiscal conditions across developed economies. Bessent's series of interventions marks an important policy inflection point.

Mark Wilson points out that Bessent's "twist," combined with other recent policy moves, conveys a core message: coordination between the Fed and the Treasury has become more explicit, and authorities will do whatever it takes to underpin nominal growth. This is not about whether a few tens of billions in Treasury buybacks can change the debt stock — it is about the policy reaction function becoming clearer: officials will choose currency depreciation before they allow the bond market to break.

The Goldman Sachs rates team has explicitly stated that while these repo operations may marginally improve long-end supply-demand dynamics, the fundamental driver of rising global long-end yields — namely, the fiscal arithmetic of developed economies — remains unchanged. However, the positioning implications of this intervention far outweigh its technical impact: it tells investors that nominal growth will be backed by policy, and those underweight "store of value" assets have already fallen behind the curve.

Wilson believes that the return of gold, bitcoin, and metals investment logic to the center of market conversation over the past ten days is a direct reflection of this policy signal.

Nominal Growth: The Macro Narrative of NVIDIA and AI Capex

The policy objective of Bessent's intervention is clear: to underpin robust US nominal growth. According to Wilson, nothing at the micro level embodies this narrative better than NVIDIA.

Following NVIDIA's strong second-quarter results, Goldman Sachs raised its 2028 revenue forecast for the company to $955 billion — an increase of $165 billion from the prior estimate, equivalent to conjuring up a company the size of Mercedes-Benz or BP (by revenue) overnight.

Even more striking, Goldman Sachs projects NVIDIA's revenue will grow by $543 billion over the next two years (from $412 billion in 2026 to $955 billion in 2028), equivalent to creating the combined size of Europe's two largest revenue generators — Volkswagen and Shell — within just two years.

Meanwhile, August saw record-high investment-grade bond issuance, with hyperscalers contributing significantly. NVIDIA's announcement of a $500 billion financing alliance to back GPU residual values carries even deeper implications: it will drive standardization of compute specifications and capacity, remove financing bottlenecks as a potential constraint on growth, and effectively lower the cost of capital for the largest builders of AI infrastructure.

The AI Narrative Shift: From "Construction" to "Application"

Two corporate announcements provide vivid illustrations of AI entering a new phase.

On August 19, Moderna, Inc. shares surged 175% in a single day — the largest one-day gain ever recorded by an S&P 500 component — following the release of research results on personalized vaccines for preventing melanoma recurrence, demonstrating the feasibility of customized, highly effective personal cancer treatment.

That same day, Stripe announced its acquisition of OpenRouter, writing in its shareholder letter: "We determined that January 1st marked the beginning of the singularity... We see a major inflection point in long-term trends... AI's promise and collective hope lies in delivering greater material prosperity and abundance."

Mark Wilson believes these two announcements clearly reveal the innovative possibilities of AI moving from the "construction" phase to the "utility" phase. AI is no longer just chatbots — Moderna, Inc.-style applications and Stripe's articulation of the "singularity" are no longer science fiction.

Market attention is shifting from the beneficiaries of the AI construction phase (memory, semiconductors, hardware) toward the true beneficiaries of the AI deployment phase — companies whose share prices have yet to reflect this transition.

Market Fractures: Two Narratives, One Market

Despite the overall positive tone of these developments, risk management in August has been far from smooth sailing.

Following NVIDIA's earnings release, its market capitalization surged by nearly $500 billion in a single day, only to give back approximately $250 billion the next day. The IGV software ETF recorded its best single-day performance in two decades. However, market breadth that day was extremely poor — this century, there has been only one other instance where the S&P 500 rose more than 60 basis points with so few constituent stocks advancing.

The "S&P 500 ex-AI" fell 0.7% on that day, and Goldman Sachs' constructed "AI winners vs. AI risk" paired portfolio declined 3%.

Mark Wilson's interpretation: NVIDIA's remarkable revenue growth is likely cannibalizing spending in other areas, just as Salesforce's results suggest that software itself will become a key beneficiary of AI in certain segments.

The difficulty of positioning and forecasting around AI exposure is undoubtedly rising — software remains heavily positioned on the short side of Goldman Sachs' 12-month momentum factor, yet carries comparable weight on the long side of the three-month momentum factor. The high correlation between AI winners and momentum factors in June and July suggests that the unwinding of consensus positions could take longer than expected, with a compression of total exposure potentially following.

In summary, Wilson believes that when all of August's events are considered together, the direction is consistent: governments will underpin nominal growth, AI capital expenditure will not pause, and the "utility" phase is no longer science fiction.

However, the market is not pricing a single story — it is pricing two stories, and those two stories do not exist on the same tape.

Record investment-grade bond issuance from hyperscalers, a $500 billion GPU residual value backstop, and the best single-day software performance in two decades — occurring simultaneously with the S&P 500 ex-AI declining and market breadth ranking among the worst this century, while the AI winners paired portfolio suffered significant losses.

Wilson's view: this is not a clean bull market. This is a market that has become convinced the state will not allow long-end yields or nominal growth to fail, but is currently debating, name by name, who will truly benefit from it.