AI tape stretches into NVDA earnings — SPY 1.5 points from 750, no fear premium, TSLA flashing the only warning

Today's tape was a single-story session dressed up as broad strength. BlackRock's iShares S&P 500 (SPY) closed +0.78% at 748.08 and Invesco QQQ Trust (QQQ) added 0.7% to 719.72, both fresh records, but the engine was almost entirely NVIDIA (NVDA) +4.4% to 235.75 on the U.S. Commerce Department's H200 approval for roughly ten Chinese buyers and a cluster of analyst target hikes (Cantor to $350, Bank of America to $320, UBS to $275). Cisco's 13% post-print pop reinforced the AI-capex story and dragged the Dow back above 50,000 for the first time since the Iran war began. That is a narrow win column.

Underneath, the rotation is starting to bifurcate. Tesla (TSLA) closed -0.5% at 443.18 with the BYD April export number (135,098 vehicles, more than Tesla sold globally) sitting on the tape and the Hardware 3 / Full Self-Driving litigation overhang building. Strategy (MSTR) +5% to 186.93 looks like risk-on confirmation, but a 13F print this week showed Jane Street cut its common-stock stake 78% quarter-over-quarter in Q1, rotating into Ether ETFs — retail is buying the spectacle while one of the better-informed prop desks de-risks. The day's read: AI infrastructure is still the only trade, and the entire market's center of gravity is now six trading days from NVDA's FQ1 print.

SPY printed a session high of 749.53 — $1.45 from the 750 round number — on 77.5M shares, which is healthy but not climactic. The Cboe Volatility Index (VIX) closed at 17.87, down a tick from 17.99, sitting squarely inside a five-session 17.08–18.38 range. Run that against 60-day realized vol on SPY at 15.6% and you get a roughly two-and-a-quarter point implied premium — modest, in line with a trending-regime tape, and frankly thin given the binary calendar risk one week out. Vol-sellers are comfortable. They probably should be a little less comfortable.

The Fed funds effective rate sits at 3.64% (April print, unchanged), confirming the cutting cycle is paused. The longer end of the curve is doing the talking instead: 10-year Treasury yields drifted to 4.47% on Thursday after a hotter-than-expected April PPI on Wednesday — the biggest monthly jump since March 2022. That should be a headwind for duration-sensitive growth, and yet QQQ and NVDA went the other way. The read is that hyperscaler capex has decoupled from the rates conversation; AI demand is treated as a non-cyclical commitment, and the Goldman note flagging the Nasdaq-100 / 1-month-call correlation at the highest level since January 2017 captures exactly that posture. The 2017 analogue is bullish until it isn't — that year was also the calmest VIX year on record, and the calm itself was the setup for what came in 2018.

NVIDIA (NVDA) +4.4% to 235.75 is the day's only story that matters. Volume printed 327M shares — well above the recent run rate — and the move broke 230 clean with the session high at 236.54. The catalyst stack is real: Commerce approved H200 sales to roughly ten Chinese buyers, Cantor Fitzgerald's C.J. Muse went to $350 citing sold-out compute into 2027 and a 10–15% buyback authorization possibility, and Wolfe Research reiterated NVDA as its top AI pick on the agentic-AI hyperscaler-capex thesis. The cautionary tail: NVDA is up roughly 20% in the past month and has fallen in three of the last four reports despite beating; FQ1 guidance is already at ~$78B (75% YoY), which is the bar to clear, not the surprise.

Strategy (MSTR) +5% to 186.93 printed an intraday high of 193 on Saylor's reiteration of the 'buy 10 to 20 BTC for every one we sell' framework and continued accumulation chatter. The bull tell is the volume — 28.8M shares, well above average. The bear tell is the Jane Street 13F: down 78% on the MSTR common stake quarter-over-quarter, with IBIT and FBTC also cut hard, and the proceeds rotated into Ether ETFs. Crowd is one direction; one of the smarter institutional books is going the other. Worth holding both thoughts.

Tesla (TSLA) -0.5% to 443.18 is the laggard worth naming. The cleantech press flagged BYD's 135K April vehicle exports as exceeding Tesla's entire global April sell-through, and Musk's admission this week that Hardware 3 cars need a discounted trade-in path to reach unsupervised FSD creates a real product-liability tail. Social sentiment scores it at 28/100, prediction markets imply 78.5% odds of sub-350K Q1 deliveries, and the stock is flat year-to-date while the index sets records. This is the only ticker in the seven where bears have specific, numeric arguments and bulls have vibes.

Semis told two stories today. NVDA and Broadcom drove the leadership, but Micron -1.5%, Intel -3.4%, and Qualcomm -6.1% all sold off. That isn't a broad semi rally — that's a quality-and-AI-pure-play bid with everything else getting trimmed. JPMorgan's semiconductor downgrade earlier this week is starting to look like the smarter call on the second-tier names even as the megacap leaders extend.

Networking and AI-adjacent enterprise gear was the real broadening signal: Cisco +13.4% on a beat and capex-spend forecast was its best day in nearly 15 years and pulled the Dow back to 50,063. Software held its ground (MSFT +1%, ORCL +3%) but did not lead. Defensives didn't confirm — bonds drifted, the curve steepened slightly on the PPI print, and the dollar firmed on safe-haven flows tied to the Trump–Xi Beijing summit and lingering Iran-Strait tensions. The cross-asset tell is that this is still a 'concentrated risk-on' tape, not a breadth thrust. When the leadership list narrows to NVDA and Cisco, you're trading a theme, not an economy.

Overnight: equity futures closed firm but not euphoric, with Asian semis (Korean Kospi +1.8% on AI exposure) likely to extend Thursday's lead. The Trump–Xi Beijing summit continues into Friday, with the Strait of Hormuz language and any read on H200/Blackwell export framework as the headline risk. Crude is roughly flat near $105 Brent, dollar index around 95.25 — neither moving aggressively but both biased toward firmer.

The calendar between now and next Wednesday's NVDA print is dominated by retail readthrough: Walmart and Home Depot report next week, with Cerebras's 65% IPO-day pop sucking attention into the AI ecosystem broadening trade. Macro data is light — Fed speakers and housing prints, no CPI or NFP in the window. The one thing that would change my view is a VIX close above 19 combined with SPY rejecting 750 on volume — that would tell you the call-skew Goldman flagged is starting to unwind, and the late-cycle 2017 analogue is closer to 2018 than people are positioned for. As Goldman's strategists put it this morning, the current Nasdaq-100/call-price correlation 'could suggest even more potential bullish action ahead' — fine, but that correlation is itself a positioning artifact, not a fundamental signal.

Base case carries the trend: SPY through 750 on the next session, NVDA holding above 230 into the print, VIX drifting in the 17–18 range. Bull case is a Cisco-style melt-up extension with SPY printing 755 and QQQ above 725, driven by H200/Blackwell headlines or a buyback pre-announcement out of NVIDIA. Bear case is the one with the asymmetric payoff: SPY rejects 750, the call-vol structure starts unwinding, and TSLA-style internal weakness spreads to second-tier semis. The Bear bucket gets a slight elevation here (28%) because the entire tape is leaning on one name with one binary catalyst six sessions out — that's not broad-based macro strength, that's idiosyncratic concentration risk.

Operating bias: stay long the AI infrastructure leaders (NVDA, AVGO, MSFT) above SPY 745, with NVDA in a trending regime where fading rallies has been the consistently wrong trade. Lean continuation above NVDA 230; the 218 level on the downside is where the structural thesis breaks and you should be flat into the print rather than long if it gets there. On SPY, a clean break above 750 on >80M shares is the green light to add; a rejection with a close back below 745 is the cue to trim into strength and let next Wednesday's print clear before re-engaging.

On MSTR, the Saylor accumulation narrative is intact but the Jane Street tape says don't chase 190; let it base. TSLA stays a no-touch into Q1 delivery confirmation — the BYD export gap and Hardware 3 litigation overhang are not in consensus numbers yet, and there's no positive catalyst until robotaxi/FSD data improves. Hedge construction matters here: at a VIX of 17.87 with 60-day realized at 15.6%, June SPY put spreads are cheap insurance against the one-event tape and worth carrying even if you stay constructive. Trim into strength if VIX closes below 16 — that would be the complacency print to fade.