By Matt Connor 2026-07-12 · Updated 2026-07-26

NVDA's May 2023 Earnings: The AI Gap

May 25, 2023: NVIDIA's guidance shocked Wall Street after the close and the stock opened 26% higher. The overnight gap that started the AI supercycle trade.

On the evening of May 24, 2023, NVIDIA guided its next quarter's revenue roughly fifty percent above Wall Street's consensus, the data-center number that made "AI capex" the market's organizing story. The stock's response happened almost entirely where most investors couldn't trade it: overnight. This page replays the gap that opened the AI supercycle era from the stored minute tape, the guidance figures themselves, whether the rest of the semiconductor complex actually moved, what the options tape printed, where the gap ranks in NVDA's own two-decade history, and the dated path to the trillion-dollar line. Every number is a stored query; expand any panel for the SQL.

What NVIDIA actually guided

The figures a searcher for this event usually wants are the guidance numbers themselves, so here they are as reported at the time. For its April quarter (fiscal Q1 2024), NVIDIA reported revenue of $7.19 billion against a consensus near $6.52 billion, with adjusted earnings of $1.09 per share versus roughly $0.92 expected, a solid beat, and not the story. The story was the outlook: revenue guidance for the July quarter of $11.00 billion, plus or minus 2%, against a Street consensus near $7.15 billion. Large companies typically guide within a few percent of expectations; this was more than fifty percent above them, on the revenue line of a company already among the world's largest, with the company pointing to data-center demand for AI chips. The guidance itself later proved conservative, the July quarter came in at $13.51 billion when NVIDIA reported it that August. Those consensus figures are press-reported context, not warehouse data; everything below is the market's tape-recorded verdict on them.

The gap, receipted

QueryNVDA: May 24, 2023 official close vs May 25 open, the overnight repricing

may24_official_close	may24_last_extended_print	after_hours_move_pct	may25_rth_open	overnight_gap_pct
305.41	380.6	24.6	385.23	26.1

The exact SQL behind every number

WITH
    (
        SELECT argMaxIf(toFloat64(close), window_start,
            (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959)
        FROM global_markets.delayed_stocks_minute_aggs
        WHERE ticker = 'NVDA'
          AND window_start >= toDateTime('2023-05-24 00:00:00') AND window_start < toDateTime('2023-05-25 04:00:00')
    ) AS rth_close_524,
    (
        SELECT argMax(toFloat64(close), window_start)
        FROM global_markets.delayed_stocks_minute_aggs
        WHERE ticker = 'NVDA'
          AND window_start >= toDateTime('2023-05-24 00:00:00') AND window_start < toDateTime('2023-05-25 04:00:00')
    ) AS last_extended_524
SELECT
    round(rth_close_524, 2) AS may24_official_close,
    round(last_extended_524, 2) AS may24_last_extended_print,
    round((last_extended_524 / rth_close_524 - 1) * 100, 1) AS after_hours_move_pct,
    round(toFloat64(argMinIf(open, window_start,
        (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959)), 2) AS may25_rth_open,
    round((toFloat64(argMinIf(open, window_start,
        (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959)) / rth_close_524 - 1) * 100, 1) AS overnight_gap_pct
FROM global_markets.delayed_stocks_minute_aggs
WHERE ticker = 'NVDA'
  AND window_start >= toDateTime('2023-05-25 04:00:00') AND window_start < toDateTime('2023-05-25 23:59:00')

Run this yourself

The official May 24 close: $305.41. Within hours, the after-hours tape had already carried the stock to $380.6, a +24.6% extended-session move on the guidance. The next morning's regular open printed $385.23: an overnight gap of +26.1%, roughly $197 billion of market value created between one closing bell and the next opening one, at the roughly 2.47 billion shares then outstanding. It remains the canonical modern example of why stocks gap: the information landed while the main market was closed, and the extended sessions repriced it before most holders could act.

The day after the gap

QueryNVDA on May 25, 2023: trading the day after the guidance

prior_close	rth_open	gap_pct	rth_low	low_et	rth_high	high_et	rth_close	day_change_pct	low_vs_prior_pct	day_shares_m	rth_minute_bars
305.41	385.23	26.1	366.35	09:38	394.8	11:27	379.79	24.4	20	151.4	390

The exact SQL behind every number

WITH
    (
        SELECT argMaxIf(toFloat64(close), window_start, (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959)
        FROM global_markets.delayed_stocks_minute_aggs
        WHERE ticker = 'NVDA'
          AND window_start >= toDateTime('2023-05-24 00:00:00') AND window_start < toDateTime('2023-05-25 04:00:00')
    ) AS prior_rth_close
SELECT
    round(prior_rth_close, 2) AS prior_close,
    round(toFloat64(argMinIf(open, window_start, (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959)), 2) AS rth_open,
    round((toFloat64(argMinIf(open, window_start, (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959)) / prior_rth_close - 1) * 100, 1) AS gap_pct,
    round(minIf(toFloat64(low), (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959), 2) AS rth_low,
    formatDateTime(toTimeZone(argMinIf(window_start, (toFloat64(low), toInt64(toUnixTimestamp(window_start))), (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959), 'America/New_York'), '%H:%i') AS low_et,
    round(maxIf(toFloat64(high), (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959), 2) AS rth_high,
    formatDateTime(toTimeZone(argMaxIf(window_start, (toFloat64(high), -toInt64(toUnixTimestamp(window_start))), (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959), 'America/New_York'), '%H:%i') AS high_et,
    round(toFloat64(argMaxIf(close, window_start, (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959)), 2) AS rth_close,
    round((toFloat64(argMaxIf(close, window_start, (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959)) / prior_rth_close - 1) * 100, 1) AS day_change_pct,
    round((minIf(toFloat64(low), (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959) / prior_rth_close - 1) * 100, 1) AS low_vs_prior_pct,
    round(toFloat64(sum(volume)) / 1e6, 1) AS day_shares_m,
    countIf((toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959) AS rth_minute_bars
FROM global_markets.delayed_stocks_minute_aggs
WHERE ticker = 'NVDA'
  AND window_start >= toDateTime('2023-05-25 04:00:00') AND window_start < toDateTime('2023-05-25 23:59:00')

Run this yourself

Here is the counter-intuitive receipt: after a +26.1% gap, the regular session itself went essentially nowhere. The open printed $385.23; a shakeout low of $366.35 hit at 09:38 ET, minutes after the bell, and even that low stood +20% above the prior close, so the gap never came close to filling. The session high of $394.8 printed at 11:27 ET, and the close, $379.79, landed within a couple percent of the open, on a colossal 151.4 million shares. The day's return belonged entirely to whoever held through the night. Buyers at the open effectively paid the full repriced level, and sellers at the open captured it. The gap was the event; the session was the settlement.

Did the whole complex move?

An "AI boom" that lived in one ticker would just be an earnings pop. The claim worth testing is that the repricing extended beyond NVIDIA the same session, so here is the same May 24-close to May 25-close measurement for a checked set of semiconductor and AI-adjacent names, with the index ETFs SPY and QQQ as market context.

QueryMay 25, 2023, close to close: the semiconductor complex around NVDA's gap

ticker	may24_close	may25_close	day_change_pct
NVDA	305.41	379.79	24.4
TSM	90.14	100.94	12
AMD	108.25	120.32	11.2
AVGO	679.33	728.8	7.3
MU	66.54	69.6	4.6
MSFT	313.85	325.96	3.9
QQQ	331.65	339.74	2.4
SPY	411.09	414.63	0.9
INTC	28.98	27.41	-5.4

The exact SQL behind every number

SELECT
    ticker,
    may24_close,
    may25_close,
    round((may25_close / may24_close - 1) * 100, 1) AS day_change_pct
FROM (
    SELECT
        ticker,
        round(argMaxIf(toFloat64(close), window_start,
            toDate(toTimeZone(window_start, 'America/New_York')) = toDate('2023-05-24') AND (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959), 2) AS may24_close,
        round(argMaxIf(toFloat64(close), window_start,
            toDate(toTimeZone(window_start, 'America/New_York')) = toDate('2023-05-25') AND (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959), 2) AS may25_close
    FROM global_markets.delayed_stocks_minute_aggs
    WHERE ticker IN ('NVDA', 'AMD', 'TSM', 'AVGO', 'MU', 'INTC', 'MSFT', 'SPY', 'QQQ')
      AND window_start >= toDateTime('2023-05-24 00:00:00') AND window_start < toDateTime('2023-05-26 00:00:00')
    GROUP BY ticker
)
ORDER BY day_change_pct DESC

Run this yourself

It moved, and unevenly. The two big chipmakers nearest the trade, TSM (+12%) and AMD (+11.2%), repriced double digits the same session with no earnings of their own that evening, and AVGO added 7.3%. The broad market barely registered it: SPY closed up just 0.9%, the Nasdaq-100 ETF QQQ +2.4%. And the dispersion inside the sector is the tell that the market was re-sorting winners rather than lifting everything with a chip on it: Intel finished the same session at -5.4%. A sector repricing with a loser inside it is a different animal from a market-wide rally, and that split, visible on day one, stayed the AI trade's signature.

The options tape: after, not before

If traders had front-run the print, it would show in the listed options. Here is NVDA's full options tape, every contract on every NVDA strike and expiry, by session: five sessions before the report, the report session, and the gap day.

QueryNVDA options: contracts, trades, call share, and premium by session

session	contracts_m	trades_k	call_share_pct	premium_bn
2023-05-17	0.87	173	58.2	0.74
2023-05-18	1.78	336	59.8	1.2
2023-05-19	1.18	227	57.9	0.72
2023-05-22	0.37	90	52	0.39
2023-05-23	0.42	95	51.5	0.4
2023-05-24	0.93	203	54.5	0.77
2023-05-25	2.75	602	59.8	4.82

The exact SQL behind every number

SELECT
    toString(toDate(toTimeZone(sip_timestamp, 'America/New_York'))) AS session,
    round(sum(size) / 1e6, 2) AS contracts_m,
    round(count() / 1e3, 0) AS trades_k,
    round(100.0 * sumIf(size, substring(ticker, 13, 1) = 'C') / sum(size), 1) AS call_share_pct,
    round(sum(size * price * 100) / 1e9, 2) AS premium_bn
FROM global_markets.options_trades
WHERE ticker LIKE 'O:NVDA2%'
  AND sip_timestamp >= toDateTime('2023-05-17 00:00:00') AND sip_timestamp < toDateTime('2023-05-26 00:00:00')
GROUP BY session
ORDER BY session

Run this yourself

The pre-print tape shows no stampede. Activity ran warm through opex week, peaking at 1.78 million contracts on 2023-05-18, the session before the May 19 monthly expiration; the two sessions just before earnings were then the quietest of the window, 0.37 and 0.42 million contracts. The report session itself printed 0.93 million. Then the guidance landed: 2.75 million contracts on the gap day, the heaviest of the window, and $4.82 billion of option premium in a single session, against $0.77 billion on the report session and roughly $0.39–$0.4 billion on the quiet days. Calls took 59.8% of gap-day contract volume. One caution on reading this: it is volume, not open interest, contracts changing hands, not net positioning, and by this measure the explosion came after the news, not ahead of it.

Where the May 2023 gap ranks in NVDA history

Was this actually NVIDIA's biggest overnight repricing, or merely its most famous? Full-history receipt: every NVDA session on the minute tape back to 2003, ranked by the size of the regular-session open over the prior official close.

QueryNVDA's ten largest overnight gaps up, 2003-present

session	prior_close	open_usd	overnight_gap_pct
2023-05-25	305.41	385.23	26.1
2003-11-07	18.18	21.66	19.1
2016-11-11	67.77	79.51	17.3
2004-11-05	15.35	17.72	15.4
2008-11-07	7.63	8.68	13.8
2023-02-23	207.65	234.4	12.9
2024-02-22	674.85	750.25	11.2
2009-12-07	14.26	15.85	11.2
2017-05-10	102.89	114.29	11.1
2011-08-12	13.41	14.83	10.6

The exact SQL behind every number

WITH daily AS (
    SELECT
        toDate(toTimeZone(window_start, 'America/New_York')) AS et_date,
        argMinIf(toFloat64(open), window_start, (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959) AS rth_open,
        argMaxIf(toFloat64(close), window_start, (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959) AS rth_close
    FROM global_markets.delayed_stocks_minute_aggs
    WHERE ticker = 'NVDA'
    GROUP BY et_date
    HAVING countIf((toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959) > 0
)
SELECT
    toString(et_date) AS session,
    round(prev_close, 2) AS prior_close,
    round(rth_open, 2) AS open_usd,
    round((rth_open / prev_close - 1) * 100, 1) AS overnight_gap_pct
FROM (
    SELECT et_date, rth_open, rth_close,
           lagInFrame(rth_close) OVER (ORDER BY et_date ASC ROWS BETWEEN 1 PRECEDING AND CURRENT ROW) AS prev_close
    FROM daily
)
WHERE prev_close > 0 AND isFinite(rth_open / prev_close)
ORDER BY overnight_gap_pct DESC
LIMIT 10

Run this yourself

May 25, 2023 is the largest overnight gap up in NVDA's recorded minute-tape history: +26.1%, well clear of the runner-up (2003-11-07, +19.1%, from a prior close of $18.18). The rest of the list is mostly mornings after earnings reports too, including 2023-02-23 and 2009-12-07, the February 2023 and February 2024 AI-era prints that each cleared ten percent and still land nowhere near this one.

The path to a trillion dollars

The gap did not fade, it compounded. At the roughly 2.47 billion shares outstanding NVIDIA reported that quarter, a $1 trillion market value worked out to about $404.86 per share, the line the financial press watched that week. Here is the dated path from the pre-gap close:

QueryNVDA daily closes from the pre-gap session through mid-June 2023

session	close_usd	change_pct	cum_vs_pregap_pct
2023-05-24	305.41	-0.5	0
2023-05-25	379.79	24.4	24.4
2023-05-26	389.24	2.5	27.4
2023-05-30	401.13	3.1	31.3
2023-05-31	378.38	-5.7	23.9
2023-06-01	397.74	5.1	30.2
2023-06-02	393.24	-1.1	28.8
2023-06-05	391.74	-0.4	28.3
2023-06-06	386.54	-1.3	26.6
2023-06-07	374.83	-3	22.7
2023-06-08	385.1	2.7	26.1
2023-06-09	387.64	0.7	26.9
2023-06-12	394.97	1.9	29.3
2023-06-13	410.23	3.9	34.3
2023-06-14	429.98	4.8	40.8
2023-06-15	426.46	-0.8	39.6
2023-06-16	427.02	0.1	39.8

The exact SQL behind every number

WITH
    daily AS (
        SELECT
            toDate(toTimeZone(window_start, 'America/New_York')) AS et_date,
            argMaxIf(toFloat64(close), window_start, (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959) AS rth_close
        FROM global_markets.delayed_stocks_minute_aggs
        WHERE ticker = 'NVDA'
          AND window_start >= toDateTime('2023-05-23 00:00:00') AND window_start < toDateTime('2023-06-17 00:00:00')
        GROUP BY et_date
    ),
    (SELECT any(rth_close) FROM daily WHERE et_date = toDate('2023-05-24')) AS pre_gap_close
SELECT
    toString(et_date) AS session,
    round(rth_close, 2) AS close_usd,
    round(if(prev_close = 0, NULL, (rth_close / prev_close - 1) * 100), 1) AS change_pct,
    round((rth_close / pre_gap_close - 1) * 100, 1) AS cum_vs_pregap_pct
FROM (
    SELECT et_date, rth_close,
           lagInFrame(rth_close) OVER (ORDER BY et_date ASC ROWS BETWEEN 1 PRECEDING AND CURRENT ROW) AS prev_close
    FROM daily
)
WHERE et_date >= toDate('2023-05-24')
ORDER BY et_date

Run this yourself

QueryThe trillion-dollar crossing, receipted: first touch and first close above $404.86

overnight_cap_added_bn	first_touch_session	first_touch_high	first_touch_close	sessions_to_first_touch	first_close_session	first_close_usd	sessions_to_first_close	first_close_cum_pct
197	2023-05-30	419.38	401.13	3	2023-06-13	410.23	13	34.3

The exact SQL behind every number

WITH
    daily AS (
        SELECT
            toDate(toTimeZone(window_start, 'America/New_York')) AS et_date,
            argMaxIf(toFloat64(close), window_start, (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959) AS rth_close,
            maxIf(toFloat64(high), (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959) AS rth_high
        FROM global_markets.delayed_stocks_minute_aggs
        WHERE ticker = 'NVDA'
          AND window_start >= toDateTime('2023-05-24 00:00:00') AND window_start < toDateTime('2023-06-20 00:00:00')
        GROUP BY et_date
    ),
    firsts AS (
        SELECT
            minIf(et_date, rth_high >= 404.86) AS ft,
            minIf(et_date, rth_close >= 404.86) AS fc
        FROM daily
    ),
    (SELECT any(rth_close) FROM daily WHERE et_date = toDate('2023-05-24')) AS pre_close,
    (
        SELECT argMinIf(toFloat64(open), window_start, (toHour(toTimeZone(window_start, 'America/New_York')) * 60 + toMinute(toTimeZone(window_start, 'America/New_York'))) BETWEEN 570 AND 959)
        FROM global_markets.delayed_stocks_minute_aggs
        WHERE ticker = 'NVDA'
          AND window_start >= toDateTime('2023-05-25 04:00:00') AND window_start < toDateTime('2023-05-26 00:00:00')
    ) AS may25_open
SELECT
    round((may25_open - pre_close) * 2.47, 0) AS overnight_cap_added_bn,
    toString(any(f.ft)) AS first_touch_session,
    round(anyIf(d.rth_high, d.et_date = f.ft), 2) AS first_touch_high,
    round(anyIf(d.rth_close, d.et_date = f.ft), 2) AS first_touch_close,
    countIf(d.et_date > toDate('2023-05-24') AND d.et_date <= f.ft) AS sessions_to_first_touch,
    toString(any(f.fc)) AS first_close_session,
    round(anyIf(d.rth_close, d.et_date = f.fc), 2) AS first_close_usd,
    countIf(d.et_date > toDate('2023-05-24') AND d.et_date <= f.fc) AS sessions_to_first_close,
    round((anyIf(d.rth_close, d.et_date = f.fc) / pre_close - 1) * 100, 1) AS first_close_cum_pct
FROM daily AS d
CROSS JOIN firsts AS f

Run this yourself

The report day itself, May 24, had closed at -0.5%, the market's last pre-AI-boom price, and the "slightly down" session the gap made famous. Then +24.4% on the gap day and +2.5% the day after. 3 sessions after the guidance, on 2023-05-30, the intraday tape ran to $419.38, through the trillion line, and closed back below it at $401.13; the next session gave back -5.7%. The first close above the line came on 2023-06-13, at $410.23, 13 trading sessions after the guidance, +34.3% above the pre-gap close. By 2023-06-14 the stock stood at $429.98, +40.8% in three weeks. All prices here are as-traded in 2023, before NVDA's 2024 ten-for-one stock split. Within two years the AI-capex trade this evening ignited had become the market's dominant theme, the same complex whose 2026 whipsaws fill our modern NVDA coverage.

What this day teaches

Earnings gaps are the purest form of the overnight problem: the reaction window belongs to the extended session's thin, fast tape, and the regular session mostly ratifies it. The receipt worth remembering from this one, the biggest single-name repricing of its era added almost nothing between its open and close. If your plan for events is "react at the regular open," this tape prices exactly what that plan is worth.

The episode also rewrote what an earnings print could be. Guidance surprises of a few percent move stocks; this was a guidance surprise of roughly fifty percent on the revenue line of a company already among the world's largest, a category error in the consensus rather than a beat, and the market treated it as new information about an entire economy's capital-spending direction rather than one company's quarter. That is why the complex moved with it, why a competitor could fall on the same news, and why the date is remembered as the AI trade's starting gun rather than as an NVDA story alone.

NVDA AI-boom FAQ

How much did NVDA gap up after May 2023 earnings?

The regular session opened +26.1% above the prior official close, $305.41 to $385.23, after the after-hours tape had already moved +24.6% on the guidance.

What did NVIDIA announce on May 24, 2023?

Fiscal Q1 2024 results, revenue of $7.19 billion against a consensus near $6.52 billion, plus the shock: July-quarter revenue guidance of $11.00 billion, plus or minus 2%, versus a Street consensus near $7.15 billion, on data-center demand for AI chips.

Was this NVIDIA's biggest earnings gap ever?

Yes, on the minute tape back to 2003: the +26.1% overnight gap into May 25, 2023 is NVDA's largest recorded gap up, ahead of the +19.1% runner-up from 2003-11-07.

When did NVIDIA first reach a $1 trillion market cap?

The tape first crossed the widely watched $404.86 threshold intraday on 2023-05-30 (high $419.38) and first closed above it on 2023-06-13 at $410.23, 13 trading sessions after the May 24 guidance.

Did other semiconductor stocks move with NVDA?

Yes, TSM added 12% and AMD +11.2% the same session, while SPY rose under one percent and Intel closed lower. It was a sector repricing, not a market-wide move.

Every panel above is a stored, versioned query over the historical tape, expand the SQL to see each measurement. Want to feel this day instead of reading it? It is one of the playable scenarios in the Strasmore Labs trading simulator.

#market history #nvda #earnings#ai boom