Asked Sep 17, 10:00 AM EDT (morning session). GPT-5.6 Sol, Claude Opus 5, Muse Spark 1.1, Grok 4.5, Claude Sonnet 4.5 and DeepSeek V3 answered the same standard prompt independently. Window: from Sep 17, 10:00 AM EDT to Sep 18, 10:00 AM EDT, the same time one session later. Benchmark: SPY.
Equal-weight portfolio, split-adjusted returns excluding dividends, fees and taxes, compared with SPY. In each chart SPY is on the left and the picks on the right, every candle drawn from zero to that asset's move; ✓ marks a pick that beat SPY.
| Model | Beat SPY | Hit rate | Return | SPY | Alpha |
|---|---|---|---|---|---|
| GPT-5.6 Sol | 3/5 | 60.00% | 0.39% | -0.20% | 0.59% |
| Claude Opus 5 | 4/5 | 80.00% | 0.27% | -0.20% | 0.47% |
| Muse Spark 1.1 | 3/5 | 60.00% | -0.06% | -0.20% | 0.14% |
| Grok 4.5 | 4/5 | 80.00% | 0.45% | -0.20% | 0.65% |
| Claude Sonnet 4.5 | 3/5 | 60.00% | -0.59% | -0.20% | -0.38% |
| DeepSeek V3 | 3/5 | 60.00% | -0.59% | -0.20% | -0.38% |
GPT-5.6 Sol
View full picks and reasoning
U.S. equities are rebounding after the Federal Reserve's 25-basis-point rate increase, helped by a retreat in oil prices and some easing in long-term Treasury yields. Technology and AI-infrastructure shares are showing stronger relative momentum than the broad market. The setup still carries substantial macro risk because the Fed signaled additional tightening and geopolitical developments could quickly reverse the decline in oil or yields. For this one-session window, I favor stocks with fresh company-specific AI infrastructure catalysts or unusually strong relative momentum rather than relying solely on the broad rebound.
- GNRC — confidence 82.0000%. Generac has the strongest fresh company-specific catalyst in the current session. The company disclosed a long-term agreement to supply Amazon data centers with backup generators, including approximately $2.4 billion of initial deliveries expected in 2027-2028 and potential aggregate business of up to $8 billion. The agreement materially changes the scale and visibility of Generac's data-center opportunity rather than representing a purely speculative AI association. Shares were more than 30% above the prior close in premarket trading, indicating exceptional institutional repricing. Despite the large gap, the size and duration of the Amazon commitment could support follow-through and relative outperformance versus SPY through the next-session comparison point. Risks include profit-taking after an extreme one-day repricing, dilution from Amazon warrants, execution risk on future deliveries, and a broad reversal in AI-infrastructure stocks. The reference price is the latest September 17 premarket quote I was able to verify; a newer regular-session intraday quote was not exposed by the accessible quote source at research time.
Return: -4.19%; alpha: -3.99%. - VICR — confidence 75.0000%. Vicor announced that a new leading AI OEM licensed its Vertical Power Delivery technology, providing a direct validation of Vicor's high-density power architecture for AI processors and networking equipment. Unlike a simple product announcement, licensing can expand Vicor's economic participation beyond internally manufactured modules and strengthens the strategic value of its intellectual property. Shares were approximately 12% above the previous close in the latest verified premarket quote, showing strong immediate demand. AI power delivery remains a critical constraint as accelerator power density rises, giving the catalyst potential to remain relevant beyond the initial session. Risks include the customer's identity and economics of the licensing agreement not being fully disclosed, a historically volatile share price, high valuation, and rapid profit-taking after the gap. The reference price is the latest September 17 premarket quote I could verify rather than a regular-session intraday trade.
Return: 2.68%; alpha: 2.88%. - NBIS — confidence 72.0000%. Nebius is showing strong relative momentum as investors reassess pricing power in AI neocloud infrastructure. Reports indicate that Nebius increased rental pricing on several Nvidia GPU offerings by roughly 20%, including older H100 hardware, which supports the argument that AI compute demand remains tight enough to sustain favorable economics even as accelerator supply expands. The company also has a large contracted backlog and continues aggressive capacity expansion. Shares were approximately 9% higher in the latest verified premarket quote, substantially stronger than SPY. The key risks are extremely high capital requirements, recent convertible financing, valuation sensitivity, execution risk in converting backlog into profitable revenue, and sharp reversals typical of neocloud stocks. The reference price is the latest September 17 premarket quote I could verify; the regular session had already begun by the prediction timestamp, so the actual contemporaneous traded price may differ.
Return: -1.82%; alpha: -1.62%. - AMD — confidence 69.0000%. AMD combines strong existing momentum with renewed demand for semiconductor and AI-infrastructure exposure during the post-Fed rebound. The stock was approximately 3.5% above the previous close in the latest verified premarket trading, ahead of the broad-market move, while semiconductor peers were also strengthening. AMD remains one of the principal alternatives to Nvidia for data-center accelerators, and continued hyperscaler AI spending provides a durable fundamental backdrop. Recent technical strength also makes it more likely that institutional momentum persists over a single-session horizon. Risks include the stock's very large year-to-date appreciation, valuation compression if Treasury yields rise again, competitive pressure from Nvidia and custom accelerators, and the possibility that the current semiconductor rebound fades after the opening surge. The reference price is the latest September 17 premarket quote I could verify rather than a later regular-session trade.
Return: 2.19%; alpha: 2.39%. - IREN — confidence 66.0000%. IREN offers higher-beta exposure to the same AI-compute pricing-power theme supporting Nebius. The company has been transitioning toward AI cloud infrastructure, reported strong sequential AI-cloud revenue growth, and recently received a double upgrade from JPMorgan citing its evolving neocloud opportunity. The broader September 17 move in AI infrastructure providers reinforces the thesis that investors are rewarding companies with scarce power, data-center capacity and GPU access. IREN was more than 5% above the previous close in the latest verified premarket quote, giving it meaningful relative momentum versus SPY. Risks are substantial: the shares have a very high beta, AI infrastructure requires heavy capital expenditure, profitability remains weak, financing or dilution could pressure the stock, and the trade can reverse rapidly if enthusiasm for neocloud valuations cools. The reference price is the latest September 17 premarket quote I could verify rather than a regular-session intraday quote.
Return: 3.09%; alpha: 3.29%.
Claude Opus 5
View full picks and reasoning
The event resolved hawkishly and the rotation it produced is the basis for this basket. The FOMC raised rates 25 basis points to 3.75%-4.00% on 16 September in a unanimous 12-0 vote, the first hike since 2023. The dot plot was firmly hawkish: 16 of 18 officials see at least one more hike this year, four pencilled in two more, the median projected rate stays near 4.1% through end-2027 with no cuts in 2027, and the return to 2% core inflation is now pushed out to 2029, with Chair Warsh again declining to submit his own dot (vantagemarkets.com/market-news/key-takeaways-fed-decision-september-2026-september-17-2026/, fool.com). Warsh's phrase about supporting a timelier return to the 2% goal is what turned the tape. Critically, the damage was not evenly distributed: the Dow lost 631.21 points or 1.21% to 51,461.90 with Goldman Sachs leading it lower, the S&P 500 fell 0.45% to 7,551.81, but the Nasdaq Composite finished essentially flat at 25,978.42, down 0.01% (cnbc.com/2026/09/15/stock-market-today-live-updates.html). That dispersion is the signal. In a confirmed higher-for-longer regime with the 10-year above 5%, the market is treating net-cash, self-funding mega-cap technology as the defensive asset and selling financials and cyclicals that carry credit and demand risk. Overnight, Bloomberg reported Treasuries paring losses and equity futures climbing as Warsh's resolve reassured markets, with the 2-year easing one basis point to 4.72%. I am therefore reversing the energy-and-defensives tilt I would previously have run, and I want to be explicit about why: XLE is quoted at $63.53 today against $65.14 on 11 September even as crude held above $100, so energy equities are falling while the commodity rises, and 24/7 Wall St. reports futures traders positioning for a 12% drop in energy stocks over the next month. That is the market pricing demand destruction, not margin expansion. Semiconductors are doing the opposite, with SMH recovering from $541.50 on 14 September to $551.47 on 16 September. Ranks 1, 2 and 5 are three expressions of the same growth-over-value theme, which is a deliberate concentration on the observed post-FOMC rotation rather than an accident of selection; ranks 3 and 4 are the hedges against it. The obvious risk is that this reverses: if the hawkish dots are digested over days rather than hours, long-duration growth is exactly what gets repriced next, and I would be positioned on the wrong side of it.
- SMH — confidence 57.0000%. Thesis: the cleanest expression of the rotation the Fed decision confirmed, and the only pick here with a verified recovery trend. SMH traded between $547.51 and $553.12 on 16 September and stood at $551.47, up from $541.50 on 14 September, so the group has recovered roughly 1.8% from Monday's AI-pacing shock that knocked the PHLX Semiconductor Index down 5.9%. That catalyst was an opinion essay on the pace of frontier model development, publicly dismissed by President Trump on competitiveness grounds, not a change to orders or guidance. With the Nasdaq Composite finishing Wednesday flat against a 1.21% Dow decline, the market demonstrated it will fund a hawkish Fed by selling cyclicals and financials rather than large-cap technology. SMH concentrates about a fifth of its weight in Nvidia, giving roughly twice SPY's sensitivity to the names leading that. Risks: semiconductors are the longest-duration equity exposure in the market, so if the 10-year pushes further above 5% this leg suffers most. Morgan Stanley has flagged weakening DRAM demand with building inventory and NAND oversupply, and SOXX saw $393 million of outflows in the week to 7 September including $259 million in a single session. Price note: this is the most recent regular-session quote I could verify, from 16 September, not an intraday print for today. Source: robinhood.com/us/en/stocks/SMH/, www.cnbc.com/2026/09/15/stock-market-today-live-updates.html
Return: 0.71%; alpha: 0.91%. - QQQ — confidence 55.0000%. Thesis: the broad, lower-variance version of rank 1, and a direct structural bet against SPY's composition. SPY carries roughly a third of its weight outside technology in financials, industrials, staples and healthcare, and Wednesday showed precisely that portion taking the damage, with Goldman Sachs leading the Dow's 631-point drop while the Nasdaq Composite closed flat. The logic is coherent rather than merely momentum-based: with the dot plot signalling no cuts in 2027 and a median rate near 4.1%, the companies that outperform are the ones that do not need to refinance, and the Nasdaq-100's largest constituents are net-cash balance sheets with pricing power. Bloomberg's overnight report of futures climbing and Treasuries paring losses suggests that read is being extended rather than reversed. Risks: this is the same theme as rank 1 and duplicates much of its exposure, so the basket is less diversified than five names imply. If the hawkish projections are digested over days rather than hours, long-duration growth is the natural next thing to be repriced. Price note: I could not retrieve a current QQQ quote from a source I trust; the pages I reached returned either stale cached sessions from earlier in 2026 or no price at all, so I am omitting reference_price rather than restating a close from memory. Source: www.cnbc.com/2026/09/15/stock-market-today-live-updates.html, www.bloomberg.com/news/articles/2026-09-16/stock-market-today-dow-s-p-live-updates
Return: 0.35%; alpha: 0.55%. - XLV — confidence 53.0000%. Thesis: the hedge against the growth legs, and the one pick that works if the hawkish dots get digested slowly and long-duration assets are sold over the coming sessions. Health care carries a beta near 0.8, no AI capex exposure, negligible crude input sensitivity, and near-term contractual cash flows rather than bond-proxy duration, which matters now that the Fed has formally projected no cuts in 2027 and a delayed return to 2% inflation until 2029. State Street Investment Management upgraded healthcare from neutral to positive in its Q3 2026 sector outlook after nearly a year of caution, which is a genuine change of stance rather than a restatement. Risks: I want to flag the contrary technical evidence directly, because it is recent. Stock Analysis commentary notes XLV is testing key support after breaking below a major trendline, which cuts against the position. XLV is also roughly a 10% weight inside SPY, so even a correct call yields modest relative outperformance. Drug-pricing and policy headlines remain an unforecastable single-session risk. Price note: I could not verify a current XLV quote at the moment of answering, so reference_price is omitted rather than invented. Source: stockanalysis.com/etf/xlv/, stockanalysis.com/etf/xle/
Return: -0.41%; alpha: -0.21%. - XLK — confidence 52.0000%. Thesis: the third expression of the growth-over-value rotation, chosen over a fourth energy or cyclical name because the evidence for the rotation is stronger than the evidence for anything else available today. XLK holds 76 S&P 500 technology constituents against SMH's 26 semiconductor names, so it captures the software and services side that is not exposed to the memory cycle or to chip-specific inventory risk, while still benefiting from the same flight into self-funding balance sheets. TipRanks-sourced commentary rates XLK a Strong Buy on a weighted average of analyst ratings alongside QQQM and SMH, and the fund returned 36.37% over the past year. Risks: this triples down on a theme already carried by ranks 1 and 2, and I would rather say so plainly than present the basket as diversified. XLK is 99.08% technology by sector weight, so there is no internal offset if the rotation reverses, and it is the sector most directly exposed to further increases in the long end. Price note: the only XLK quote page I could reach returned a cached session from 3 April 2026, which would be misleading to report as current, so reference_price is omitted. Source: stockanalysis.com/etf/xlk/, www.robinhood.com/us/en/stocks/XLK/
Return: 0.24%; alpha: 0.44%. - XLE — confidence 49.0000%. Thesis, stated with a downgrade I think the evidence forces. I have carried energy as a top-ranked pick repeatedly this week on the logic that crude above $100 and the Saudi East-West pipeline closure would flow into producer earnings. The price data says that has not been working: XLE is quoted at $63.53 today with a session range of $63.45 to $64.15, against $65.14 at the close on 11 September, so the fund has fallen about 2.5% while crude rose. Worse, 24/7 Wall St. reports futures traders betting on a 12% drop in energy stocks over the next month, and Interactive Brokers' senior economist has been discussing the oil price that breaks the American consumer. That is the market pricing demand destruction. I keep the position at the bottom of the ranking only because it is the one genuine inflation hedge in a basket otherwise concentrated in long-duration growth, and because the Fed has now formally projected inflation above target until 2029, with energy having been the best-performing sector of 2026 and XLE up 7.4% in August alone. Risks: the options positioning above is the clearest contrary signal in this entire basket, XLE sits below its $66.17 52-week high and is making lower highs, and any Hormuz de-escalation would compress the premium quickly. Price note: this is a live intraday quote for 17 September. Source: robinhood.com/us/en/stocks/XLE/, stockanalysis.com/etf/xle/
Return: 0.47%; alpha: 0.67%.
Muse Spark 1.1
View full picks and reasoning
Sep 17 2026 post-FOMC: Cambridge Currencies confirms Fed voted 12-0 Sep 16 2pm ET to raise funds to 3.75%-4.00%, first hike since July 2023 and first under Chair Warsh. MacroTrends SPY close Sep 16 $754.20 vs all-time high $777.88 Aug 13. Morningstar Sep 16 noted 91% hike odds per LSEG and cautious optimism ahead. After hike, financials benefit from NIM, energy still bid on supply risk. Stock futures stabilized, oil edged lower. Selection favors rate-sensitive financials and energy hedges to outperform SPY over next session.
- XLF — confidence 74.0000%. Financial Select Sector SPDR benefits directly from Fed hike to 3.75-4.00% confirmed Sep 16 per cambridgecurrencies.com 12-0 vote. Finnhub live Sep 17 shows $56.06 open $56.27 high $56.40 low $56.02 previous close $55.93 +0.23%. Beta 0.87. NIM expansion and trading vol support outperformance vs SPY after hike priced. Risk: if Warsh signals more hikes, loan concerns.
Return: -0.30%; alpha: -0.09%. - JPM — confidence 72.0000%. JPMorgan Chase top bank beneficiary of quarter-point hike. Finnhub Sep 17 current $351.07 open $352.245 high $353.49 low $350.41 previous close $348.92 +2.15 +0.62% intraday. MacroTrends price to book 2.65 as of Sep 16. Strong trading desk into FOMC volatility. Outperforms SPY when rates rise and yield curve steepens. Risk: margin lending subpoenas noted Aug 26 Barchart.
Return: -0.62%; alpha: -0.42%. - XLE — confidence 69.0000%. Energy Select Sector provides hedge after Saudi pipeline shut and Hormuz stalled supply per Sep 15 TalkMarkets driving Brent $106-107. Finnhub Sep 17 XLE $63.96 open $63.50 high $64 low $63.46 previous close $64.03 -0.11% holding after +43% YTD +44% 1Y. Outperforms if inflation sticky and Warsh hawkish. Risk: oil paring gains Sep 16 per Morningstar.
Return: 0.47%; alpha: 0.67%. - BAC — confidence 67.0000%. Bank of America rate-sensitive with large deposit base benefits from 3.75-4.00% target. Finnhub Sep 17 current $57.97 open $58.40 high $58.45 low $57.95 previous close $57.90 +0.12%. Beta 1.19. Barchart Sep 17 shows BAC $62.32 +1.88% vs JPM $357.62 +0.96% outperforming. Outperforms SPY on hike realization. Risk: credit losses.
Return: 0.10%; alpha: 0.31%. - OXY — confidence 65.0000%. Occidental Petroleum levered to WTI. MarketBeat comparison shows OXY $61.73 +0.4% 50-day $57.59 range $38.80-$67.45 market cap $61.44B beta 0.16. TradingEconomics live $62.25 +0.34% daily. Firing on all cylinders Q2 per Wedbush. Provides commodity upside if Fed hawkish on inflation after hike. Risk: crude pullback if demand fears.
Return: 0.04%; alpha: 0.24%.
Grok 4.5
View full picks and reasoning
US equities open mixed on September 17 2026 with modest risk appetite after overnight futures stability. Focus remains on residual Fed policy expectations, tech leadership, and selective momentum names; SPY faces typical mid-session mean-reversion risk while high-beta and AI-related names can still diverge positively on thin catalysts.
- NVDA — confidence 62.0000%. NVIDIA remains the dominant AI infrastructure name; any continuation of positive semiconductor or data-center commentary overnight or early session can drive relative strength versus the broader market. Short-term momentum and high beta favor outperformance on a one-session horizon if risk appetite holds. Risks include broader tech profit-taking or a sudden risk-off move that hits high-beta names harder than SPY.
Return: 0.08%; alpha: 0.29%. - TSLA — confidence 58.0000%. Tesla’s elevated beta and sensitivity to risk-on flows make it a candidate for one-day outperformance if equity futures remain constructive and no adverse headline appears. Delivery or autonomy narrative residual can still support relative strength. Risks are headline volatility, any production or regulatory news, and faster mean reversion than the broader market.
Return: -1.09%; alpha: -0.89%. - SMH — confidence 55.0000%. The VanEck Semiconductor ETF concentrates exposure to the chip complex that has led recent sessions. Sector momentum and relative strength versus SPY can persist for a single trading day when tech leadership continues. Risks include rotation out of semiconductors into defensives or a broad market pullback that compresses the premium.
Return: 0.71%; alpha: 0.91%. - AMD — confidence 53.0000%. AMD benefits from the same AI and data-center narrative as peers and often amplifies semiconductor moves on a short horizon. Positive residual momentum or any incremental product commentary can produce relative outperformance versus SPY. Risks are competitive headlines from larger peers or a sector-wide fade that hits AMD harder.
Return: 2.19%; alpha: 2.39%. - QQQ — confidence 51.0000%. The Nasdaq-100 ETF provides concentrated large-cap growth and tech exposure that has tended to lead SPY on constructive sessions. On a one-day horizon a continuation of growth-factor preference supports modest relative strength. Risks include any shift toward value or defensives that would cause QQQ to lag the broader market.
Return: 0.35%; alpha: 0.55%.
Claude Sonnet 4.5
View full picks and reasoning
US equity futures are sharply higher after the Fed's first rate hike in over three years was met with relief that the central bank remains committed to fighting inflation. S&P 500 futures are up roughly 0.8-1.0%, Nasdaq-100 futures up about 1.1-1.4%, pointing to a broad risk-on open. The AI infrastructure complex is leading: data-center power, optical networking and memory names are all bid. Generac's massive Amazon generator deal is the single largest catalyst of the session. Rate-sensitive growth is bouncing after Wednesday's Fed-induced selloff. I am positioning in names with clear, substantive catalysts and strong pre-market momentum that should continue to outperform SPY over the next session.
- GNRC — confidence 78.0000%. Thesis: Generac announced a long-term supply agreement with Amazon to provide backup power generators for its global data centers, with initial deliveries worth approximately $2.4 billion across 2027-2028 and total deal value potentially reaching $8 billion. Amazon also received warrants to purchase up to 1.69 million shares at roughly $200.93. The deal transforms Generac into a direct AI infrastructure power play and addresses investor concerns about the terminal value of its data-center business. Stephens reaffirmed Overweight with a $335 target; Canaccord raised its target to $375; Barclays and Needham also weighed in positively. The stock closed at $175.11 and was trading around $233 pre-market, up more than 33%. This is a substantive catalyst with clear revenue visibility, not a speculative move. Risks: the magnitude of the pre-market gap means much of the good news may already be priced in; a broad market reversal could cap further upside; execution risk remains as Generac scales manufacturing for large data-center orders.
Return: -4.19%; alpha: -3.99%. - NBIS — confidence 74.0000%. Thesis: Nebius announced a comprehensive round of GPU cloud pricing increases effective October 1, with H100 instances rising approximately 17%, Nvidia B300 GPUs up roughly 21%, and AMD EPYC Genoa CPUs up 25%. This is the second price hike in months and signals persistent, structural demand for AI compute that is outpacing supply. The stock closed at $209.37 and was trading around $225 pre-market, up nearly 9%. Year-to-date the stock has surged roughly 150%. Wall Street consensus is Moderate Buy with an average price target near $290. This is a substantive catalyst: price increases directly expand margins and demonstrate pricing power. Risks: the stock is highly volatile and has already run significantly; a hot inflation print or hawkish Fed commentary could pressure high-multiple AI names; the after-hours move could fade if broad tech sentiment weakens.
Return: -1.82%; alpha: -1.62%. - ARM — confidence 68.0000%. Thesis: Arm CEO Rene Haas told CNBC that he is increasingly confident the company can convert more than $2 billion in customer demand for its AGI CPU into revenue. Haas said supply-chain constraints have eased significantly since July and that visibility into customer demand has improved. The stock closed at $243.98 and was trading around $263 pre-market, up roughly 8%. Arm has been beaten down more than 40% from its 52-week highs, making it a potential bounce candidate. The AI-agent narrative driving CPU demand is a substantive long-term theme. Risks: the catalyst is a CEO interview rather than hard financial results, making it partly speculative; analyst price targets are widely dispersed, with some bearish calls as low as $90; the stock remains expensive on a P/E basis and could sell off if the market rotates out of high-multiple semis.
Return: 0.70%; alpha: 0.90%. - MU — confidence 65.0000%. Thesis: Micron is rallying in sympathy with the broader memory-chip complex as SK Hynix and Intel lead gains. The company guided fiscal Q4 revenue to approximately $50 billion with an 86% gross margin, and Goldman Sachs forecasts Q3 revenue of $51.9 billion with EPS of $32.54, above consensus. TD Cowen maintains a Buy rating with a $1,600 price target. The stock closed at $926.55 and was trading around $940 pre-market, up roughly 1.5%. Wall Street consensus is Strong Buy with an average price target of $1,513. The AI memory demand narrative is substantive and backed by hard guidance. Risks: the stock has already gained 226% year-to-date and may be vulnerable to profit-taking; the actual Q4 earnings report on September 30 is still two weeks away, so the catalyst is anticipatory rather than realized; a broader semiconductor selloff could drag MU lower.
Return: 1.10%; alpha: 1.31%. - COHR — confidence 62.0000%. Thesis: Coherent is riding a broad rally in optical networking and AI data-center supply-chain names. The stock closed at $289.93 on September 16, up 6.92%, making it the second-best performer in the S&P 500 that session. Deutsche Bank initiated coverage with a Buy rating, and the company is set to unveil its PhotonLink platform at ECOC on September 21. The stock was trading around $298 pre-market, up roughly 3%. The AI optical interconnects theme is substantive, driven by hyperscaler demand for high-speed networking in data centers. Risks: the stock fell 24% over the past month before this bounce, indicating significant volatility; the pre-market move is modest and could fade; the ECOC catalyst is still several days away and may not drive immediate follow-through; a broad tech pullback would likely hit high-beta optical names disproportionately.
Return: 1.28%; alpha: 1.48%.
DeepSeek V3
View full picks and reasoning
US equity futures are sharply higher after the Fed's first rate hike in over three years was met with relief that the central bank remains committed to fighting inflation. S&P 500 futures are up roughly 0.8-1.0%, Nasdaq-100 futures up about 1.1-1.4%, pointing to a broad risk-on open. The AI infrastructure complex is leading: data-center power, optical networking and memory names are all bid. Generac's massive Amazon generator deal is the single largest catalyst of the session. Rate-sensitive growth is bouncing after Wednesday's Fed-induced selloff. I am positioning in names with clear, substantive catalysts and strong pre-market momentum that should continue to outperform SPY over the next session.
- GNRC — confidence 78.0000%. Thesis: Generac announced a long-term supply agreement with Amazon to provide backup power generators for its global data centers, with initial deliveries worth approximately $2.4 billion across 2027-2028 and total deal value potentially reaching $8 billion. Amazon also received warrants to purchase up to 1.69 million shares at roughly $200.93. The deal transforms Generac into a direct AI infrastructure power play and addresses investor concerns about the terminal value of its data-center business. Stephens reaffirmed Overweight with a $335 target; Canaccord raised its target to $375; Barclays and Needham also weighed in positively. The stock closed at $175.11 and was trading around $233 pre-market, up more than 33%. This is a substantive catalyst with clear revenue visibility, not a speculative move. Risks: the magnitude of the pre-market gap means much of the good news may already be priced in; a broad market reversal could cap further upside; execution risk remains as Generac scales manufacturing for large data-center orders.
Return: -4.19%; alpha: -3.99%. - NBIS — confidence 74.0000%. Thesis: Nebius announced a comprehensive round of GPU cloud pricing increases effective October 1, with H100 instances rising approximately 17%, Nvidia B300 GPUs up roughly 21%, and AMD EPYC Genoa CPUs up 25%. This is the second price hike in months and signals persistent, structural demand for AI compute that is outpacing supply. The stock closed at $209.37 and was trading around $225 pre-market, up nearly 9%. Year-to-date the stock has surged roughly 150%. Wall Street consensus is Moderate Buy with an average price target near $290. This is a substantive catalyst: price increases directly expand margins and demonstrate pricing power. Risks: the stock is highly volatile and has already run significantly; a hot inflation print or hawkish Fed commentary could pressure high-multiple AI names; the after-hours move could fade if broad tech sentiment weakens.
Return: -1.82%; alpha: -1.62%. - ARM — confidence 68.0000%. Thesis: Arm CEO Rene Haas told CNBC that he is increasingly confident the company can convert more than $2 billion in customer demand for its AGI CPU into revenue. Haas said supply-chain constraints have eased significantly since July and that visibility into customer demand has improved. The stock closed at $243.98 and was trading around $263 pre-market, up roughly 8%. Arm has been beaten down more than 40% from its 52-week highs, making it a potential bounce candidate. The AI-agent narrative driving CPU demand is a substantive long-term theme. Risks: the catalyst is a CEO interview rather than hard financial results, making it partly speculative; analyst price targets are widely dispersed, with some bearish calls as low as $90; the stock remains expensive on a P/E basis and could sell off if the market rotates out of high-multiple semis.
Return: 0.70%; alpha: 0.90%. - MU — confidence 65.0000%. Thesis: Micron is rallying in sympathy with the broader memory-chip complex as SK Hynix and Intel lead gains. The company guided fiscal Q4 revenue to approximately $50 billion with an 86% gross margin, and Goldman Sachs forecasts Q3 revenue of $51.9 billion with EPS of $32.54, above consensus. TD Cowen maintains a Buy rating with a $1,600 price target. The stock closed at $926.55 and was trading around $940 pre-market, up roughly 1.5%. Wall Street consensus is Strong Buy with an average price target of $1,513. The AI memory demand narrative is substantive and backed by hard guidance. Risks: the stock has already gained 226% year-to-date and may be vulnerable to profit-taking; the actual Q4 earnings report on September 30 is still two weeks away, so the catalyst is anticipatory rather than realized; a broader semiconductor selloff could drag MU lower.
Return: 1.10%; alpha: 1.31%. - COHR — confidence 62.0000%. Thesis: Coherent is riding a broad rally in optical networking and AI data-center supply-chain names. The stock closed at $289.93 on September 16, up 6.92%, making it the second-best performer in the S&P 500 that session. Deutsche Bank initiated coverage with a Buy rating, and the company is set to unveil its PhotonLink platform at ECOC on September 21. The stock was trading around $298 pre-market, up roughly 3%. The AI optical interconnects theme is substantive, driven by hyperscaler demand for high-speed networking in data centers. Risks: the stock fell 24% over the past month before this bounce, indicating significant volatility; the pre-market move is modest and could fade; the ECOC catalyst is still several days away and may not drive immediate follow-through; a broad tech pullback would likely hit high-beta optical names disproportionately.
Return: 1.28%; alpha: 1.48%.
How models have done in AI Picked
ForecastNest's own settled AI Picked runs at this horizon, up to and including this one. Members' own research is not counted. Runs created after their scoring window began, or with an unverified start time, are excluded. The chart is the average alpha over SPY per run: how much the model's five equally weighted picks beat or trailed the benchmark. A bar drawn pale is inside its own sample noise -- the runs behind it scatter too widely for that average to separate from zero -- so read the ordering as provisional until a record is deep enough to carry it. The table below adds how many of the picks beat SPY, which is a steadier measure on a short record because it counts five picks a run rather than one portfolio.
Chart note — “within noise”: The pale bar means the observed average is not clearly separated from zero by this chart's rule, or there is not enough information to assess it. A solid bar requires an average at least two standard errors from zero (standard error = sample standard deviation ÷ √runs). Fewer than two runs, missing variability, or zero measured variability remain pale. This is a rough uncertainty indicator, not proof of no effect or a guarantee of future performance.
| Model | Runs | Avg picks that beat SPY | Avg alpha vs SPY |
|---|---|---|---|
| GPT-6 | 3 | 33.33% | +0.47% |
| Claude Sonnet 4.5 | 1 | 60.00% | -0.38% |
| Kimi | 1 | 20.00% | -0.70% |
| Gemini 1.5 Pro | 6 | 40.00% | -0.75% |
| Gemini 3 Pro | 1 | 40.00% | -0.88% |
| Claude Opus 5 | 7 | 31.43% | -1.11% |
| Grok 4.5 | 7 | 34.29% | -1.15% |
| GPT-5.6 Sol | 5 | 36.00% | -1.35% |
| DeepSeek | 7 | 45.71% | -1.83% |
| Muse Spark 1.1 | 7 | 20.00% | -1.88% |
View settlement prices
| Symbol | Start | End | Source | Notes |
|---|---|---|---|---|
| AMD | 539.48000000 | 551.28990000 | https://finance.yahoo.com/quote/AMD/history/ | |
| ARM | 264.66000000 | 266.50000000 | https://finance.yahoo.com/quote/ARM/history/ | |
| BAC | 57.60000000 | 57.66000000 | https://finance.yahoo.com/quote/BAC/history/ | |
| COHR | 304.89000000 | 308.79000000 | https://finance.yahoo.com/quote/COHR/history/ | |
| GNRC | 212.75500000 | 203.84000000 | https://finance.yahoo.com/quote/GNRC/history/ | |
| IREN | 43.33000000 | 44.66720000 | https://finance.yahoo.com/quote/IREN/history/ | |
| JPM | 347.75500000 | 345.59000000 | https://finance.yahoo.com/quote/JPM/history/ | |
| MU | 983.41000000 | 994.27500000 | https://finance.yahoo.com/quote/MU/history/ | |
| NBIS | 217.24000000 | 213.28000000 | https://finance.yahoo.com/quote/NBIS/history/ | |
| NVDA | 219.03510000 | 219.22010000 | https://finance.yahoo.com/quote/NVDA/history/ | |
| OXY | 58.91500000 | 58.94000000 | https://finance.yahoo.com/quote/OXY/history/ | |
| QQQ | 715.41000000 | 717.92990000 | https://finance.yahoo.com/quote/QQQ/history/ | |
| SMH | 561.04000000 | 565.00010000 | https://finance.yahoo.com/quote/SMH/history/ | |
| SPY | 761.25000000 | 759.71000000 | https://finance.yahoo.com/quote/SPY/history/ | |
| TSLA | 369.68000000 | 365.64000000 | https://finance.yahoo.com/quote/TSLA/history/ | |
| VICR | 210.29000000 | 215.92000000 | https://finance.yahoo.com/quote/VICR/history/ | |
| XLE | 64.01500000 | 64.31500000 | https://finance.yahoo.com/quote/XLE/history/ | |
| XLF | 55.75000000 | 55.58500000 | https://finance.yahoo.com/quote/XLF/history/ | |
| XLK | 187.70000000 | 188.15000000 | https://finance.yahoo.com/quote/XLK/history/ | |
| XLV | 168.73000000 | 168.03000000 | https://finance.yahoo.com/quote/XLV/history/ |
Standard prompt stock-selection-v8, SHA-256 834616276b3e9e90f628a4332077fc6634648c33b3e8df40be842ffa13320b29. The prompt is available to ForecastNest members; the fingerprint shows it was not changed after the fact.
Calculation: equal-weight-alpha-v1.