Is Nvidia a Buy Before August 26 Earnings? Yes, With One Caveat That Decides the Sizing

TL;DR

• Yes on the business, no as an earnings trade. The distinction decides everything about how the position gets built.
• The stock sits at $223.96 (August 7 close), up from about $197 earlier this summer, when a flat 2026 had compressed the multiple toward the market average and the house call here was that boredom, not fear, was doing the mispricing.
• The print itself is a hurdle, not a catalyst: the Street already sits at or above the top of the $91 billion guide, and recent quarters have delivered records that got sold anyway.
• The caveat: the October-quarter guide and the 75% margin against rising memory costs are genuine risks, and either can produce a 5%+ air pocket in an otherwise intact thesis.

Is Nvidia a Buy Before Earnings?

Yes, at a size that treats August 26 as a risk rather than a reason. The buy case is the 2026 setup: earnings grew all year while the price went sideways, converting the most important company in the AI buildout from a story stock into something near a market multiple. That case does not need Wednesday's print to be true, and it survives a bad reaction to it.

What I would not do is buy it because earnings are coming. The full preview walks through why: consensus is parked at or above the top of management's own $91 billion guide, so the beat is pre-paid, and the last several reactions have been muted or negative on strong numbers. Buying two weeks before a print the market has already graded is how a right thesis produces a losing entry.

The Bull Case at $224

• The pie keeps growing faster than the competition arrives. The 2026 semiconductor forecast got upgraded to $1.3 trillion, memory is physically short because AI servers eat it, and every capex report this season raised the number. Nvidia sells to all of it.
• The multiple already de-rated. The compression happened without a crash: at $197 the forward multiple sat near the S&P average, an entry Goldman called compelling at the time. The stock has since recovered to $223.96, so that specific bargain is partly spent, but nothing like the 72x of the mania years is being asked.
• Zero China in the guide. Any China data-centre revenue at all arrives as upside against a bar that excluded it.

The Bear Case Deserves the Airtime

• Custom silicon is real attrition. Every hyperscaler runs an in-house chip program, and each one that ships erodes monopoly pricing a little. This is the slow argument, not the August argument.
• The margin is the August argument. Nvidia guides 75% gross margin into the memory-cost squeeze that has repriced everyone downstream. The risk is not this quarter's margin; it is the January-quarter guidance where new HBM contract pricing lands.
• Expectations do the damage, not results. A company that beats and drops is a company priced past its own guide. That is the current configuration.

How I'd Actually Do It

The position I want is shares, sized so that a 5-8% earnings air pocket is an annoyance rather than a decision. Recent earnings moves have run near 5.4%; the decade average is closer to 8%. Anyone whose size would force a sale into that kind of gap has the wrong size, not the wrong stock.

The sequencing matters more than the timing: a partial position before the print, with the rest reserved for the reaction. If the quarter is strong and the stock sells off anyway, that is the pattern of the last year, and it has been the entry, not the exit. What changes my mind is specific: a cut to the margin guide, or an October guide that breaks the sequential staircase. Either of those is new information, and the thesis reloads from scratch.

The One-Line Read

Nvidia at $223.96 is a buy on the business and a coin flip on the print, so the position gets sized for a 5-8% gap, half held back for the reaction, with the thesis pinned to the October guide and the margin line rather than to whether $91 billion gets cleared, because it will, and the market has already marked it.