Nvidia and AI technology and Stock Market

Nvidia and AI technology and Stock Market

• 1. EYE ON THE MARKET • MICHAEL CEMBALEST • J.P. MORGAN • September 3, 2024 2024 energy paper / US inflation/labor monitor / US Federal debt monitor / US economic monitor 1 A severe case of COVIDIA: prognosis for an AI-driven US equity market NVIDIA and its GPU customers are now a large driver of equity market returns, earnings growth, earnings revisions, industrial production and capital spending. NVIDIA’s financial results are extraordinary (it beat on revenues and earnings again last week, and announced a $50 billion stock buyback). As shown below, NVIDIA has also experienced the fastest road to being the market’s largest stock in the post-war era. But for investors, the more important questionslook past the economics of selling GPUs and focus on the ability of hyperscalers (Google, Amazon, Microsoft, Meta, etc) and other AI infrastructure users/providers to earn adequate returns on hundreds of billions in AI-related capital spending. The level of this spending now rivals the mainframe era of the late 1960’s and the fiber optic deployment of the late 1990’s. For adequate returns on AI infrastructure to materialize, within the next 12-18 months we will need to see a greater shift in favor of “inference” tasks (AI used to run productionmodels for corporate customers) rather than GPU capacity primarily being used to train foundational models and chatbots. In this Eye on the Market, we take a closer look. Also: some brief comments on the Third Circuit’s decision to strike down Section 230 protections for certain social media company activities. Michael Cembalest JP Morgan Asset Management Table of contents COVIDIA: GPU orbit now driving equity markets and capital spending ........................................................................... 2 NVIDIA financials: phenomenal, and the antithesis of the dot-com era.......................................................................... 4 From A (Altman) to Y (Combinator): the transformational view of generative AI........................................................... 5 And yet…there are a lot of questions on spending, revenues, productivity benefits, energy and more ........................ 6 Tracking corporate sector AI adoption so far: mostly positive but some mixed signals.................................................. 8 What are equity markets pricing in? ................................................................................................................................ 9 COVIDIA prognosis: judgment day on AI capital spending is probably 2-3 years away .................................................10 Appendix I: Current and prospective NVIDIA competitors.............................................................................................11 Appendix II: The fate of prior equity market leaders.....................................................................................................12 0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10% 20 15 10 5 0 Quarters before stock becomes largest share of market cap GM 1963 AAPL 2023 IBM 1978 MSFT 2024 NVDA 2024 XOM 2008 GE 2001 CSCO 2000 MO 1992 NVIDIA: fastest road to the top Share of market cap Source: Bloomberg, Factset, GS, JPMAM, August 2024

• 2. EYE ON THE MARKET • MICHAEL CEMBALEST • J.P. MORGAN • September 3, 2024 2024 energy paper / US inflation/labor monitor / US Federal debt monitor / US economic monitor 2 COVIDIA: GPU orbit now driving equity markets and capital spending NVIDIA and its GPU customers are an increasingly large driver of equity market returns, earnings growth, earnings revisions, industrial production and capital spending. You have probably seen the first chart before on the Mag 7 share of market cap. The others are just as notable, highlighting how earnings, industrial production and capital spending are increasingly impacted by AI related spending. Some colleagues pointed out recently that Mag 7 capital spending now exceeds that of the entire energy sector. 10% 15% 20% 25% 30% 1990 1995 2000 2005 2010 2015 2020 2025 Magnificent 7 domination % of total large-cap market cap Source: Factset, JPMAM. August 27, 2024 Mag 7: AMZN, AAPL, GOOG, META, MSFT, NVDA, TSLA -20% -10% 0% 10% 20% 30% 40% 50% S&P 1 - 20 S&P 21 - 100 S&P 101 - 200 S&P 201 - 300 S&P 301 - 500 Trailing 12 month performance Change in 2024 earnings estimate since June 2023 Top quintile of stocks are driving earnings revisions and market returns, Percent Source: J.P Morgan Equity Macro Research, June 27, 2024 60 70 80 90 100 110 120 130 140 150 160 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 US industrial production of high tech industries Index (100 = January 2020) Source: Federal Reserve Board, JPMAM, July 2024 Communication equipment Computer equipment Semiconductors All industry ex-high tech 0 100 200 300 400 500 600 700 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 Bloomberg Magnificent 7 Index Bloomberg US large cap ex Magnificent 7 Index Capex + R&D, Magnificent 7 vs rest of market Index (100 = January 2020) Source: Bloomberg, JPMAM, August 2024

• 3. EYE ON THE MARKET • MICHAEL CEMBALEST • J.P. MORGAN • September 3, 2024 2024 energy paper / US inflation/labor monitor / US Federal debt monitor / US economic monitor 3 US corporate concentration has been steadily increasing for the last 100 years; see the chart on the share of assets and revenues controlled by the largest 1% of companies. But we have now hit new extremes, exemplified by the collapse in the share of companies outperforming the S&P 500, and the outperformance of the market cap weighted S&P 500 vs the equal weighted version of the same index (last reached in the late 1990’s). 60% 65% 70% 75% 80% 85% 90% 95% 100% 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 Assets Revenues Largest 1% of US corporations share of assets and revenues, Percent Source: "100 years of corporate concentration", Kwon et al., August 2024 20% 25% 30% 35% 40% 45% 50% 55% 60% 65% 1990199219941996199820002002200420062008 20102012201420162018 202020222024 Constituents outperforming the S&P 500 Source: Bloomberg, JPMAM, August 16, 2024 Percent of companies outperforming (total return) -50% -40% -30% -20% -10% 0% 10% 20% 30% 40% 1993 1996 1999 2002 2005 2008 2011 2014 2017 2020 2023 Rolling 2 year performance of market cap vs equal weighted S&P 500 Source: Bloomberg, JPMAM, August 2024 Market cap weighted S&P 500 outperforms Equal weighted S&P 500 outperforms Third Circuit removes some Section 230 protections from social media companies • This year, in Moody v. NetChoice LLC, the Supreme Court ruled that the use of algorithms by social media or internet companies to automatically prioritize or otherwise curate content that is posted by third parties reflects that company’s editorial judgment and constitutes “expressive activity” which is protected by the First Amendment. In other words, your TikTok, Twitter or LinkedIn feed constitutes that company’s own speech, even though the content was originally created and uploaded by third parties. So…the Court essentially concluded that social media companies have a First Amendment right to use algorithms to boost or downplay posts based on whatever political, social or economic standards they wish. The decision implied that a recently passed law in Texas may be unconstitutional, and remanded the case back to lower Texas courts • Fine, said the Third Circuit: if curated feeds are first-party speech, then Section 230 of the Communications Decency Act does not shield these companies from liability for harm caused by content that their algorithms recommend to users. Thus, in Anderson v. TikTok, the court allowed a lawsuit against TikTok to proceed. Its algorithm is alleged by plaintiffs to have recommended a video featuring the “blackout challenge” to a child who died while performing it • This all makes perfect sense to me; let the lawsuits begin

• 4. EYE ON THE MARKET • MICHAEL CEMBALEST • J.P. MORGAN • September 3, 2024 2024 energy paper / US inflation/labor monitor / US Federal debt monitor / US economic monitor 4 NVIDIA financials: phenomenal, and the antithesis of the dot-com era • NVIDIA’s operating margins are high and rising. The company has seen a step-change in projected revenue growth over the last year (third chart, from gold dots to red dots). Last week, the company beat expectations on revenues and earnings, and announced a $50 billion stock buyback program • NVIDIA bears no resemblance to dot-com market leaders like Cisco (see below) whose P/E multiple also soared but without earnings to go with it • Consensus analyst reports project that NVIDIA will maintain 90%+ market share in AI chips for the next two years. Even with a possible delay in NVIDIA’s next generation “Blackwell” chip shipments, few if any firms are positioned to take immediate advantage…but competition is coming. See Appendix I on current and prospective NVIDIA competitors, and Appendix II for history since the 1960’s on what happened to stocks once they represented the largest market cap, as NVIDIA did for one brief day in June 2024 0% 10% 20% 30% 40% 50% 60% 70% NVIDIA Microsoft META Alphabet Apple Amazon Tesla Magnificent 7 operating margin comparison Percent Source: Bloomberg, JPMAM, July 2024 -20% -10% 0% 10% 20% 30% 40% 50% 60% 2009 2011 2013 2015 2017 2019 2021 2023 NVIDIA net income margins Percent Source: Bloomberg, JPMAM, July 2024 21% -7% 53% 61% 0% 126% 11% 24% 21% 18% 104% 42% 16% $0 $20 $40 $60 $80 $100 $120 $140 $160 $180 $200 $220 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 Nvidia sales growth: actuals and estimate revisions US$, billions with percent y/y growth Source: Bloomberg, JPMAM, August 30, 2024 Actual Aug 2024 projections Mar 2023 projections 2018 2020 2022 2024 0x 20x 40x 60x 80x 100x 120x 140x 1996 1998 2000 2002 CSCO NVDA P/E ratios, dot-com era vs today Forward P/E ratio Source: Factset, JPMAM, August 30, 2024 0 500 1000 1500 2000 2500 3000 2018 2020 2022 2024 1-yr Fwd EPS Share price Nvidia share price vs 1-year forward EPS Index (100 = January 2018) Source: Factset, Bloomberg, JPMAM, August 30, 2024 0 200 400 600 800 1000 1200 1400 1600 1800 2000 1996 1998 2000 2002 Share price 1-yr Fwd EPS Cisco share price vs 1-year forward EPS Index (100 = January 1996) Source: Factset, Bloomberg, JPMAM, June 25, 2024

• 5. EYE ON THE MARKET • MICHAEL CEMBALEST • J.P. MORGAN • September 3, 2024 2024 energy paper / US inflation/labor monitor / US Federal debt monitor / US economic monitor 5 From A (Altman) to Y (Combinator): the transformational view of generative AI The prognosis for COVIDIA is less related to NVIDIA financials and more related to whether/when the AI transformation markets are pricing in will actually happen. AI adherents make powerful arguments: • Sam Altman at OpenAI considers AI the “biggest, best and most important of all technology revolutions”, and believes it will become more ubiquitous over time0F 1 • According to McKinsey (never the types to shy away from seismic predictions against which they’re never measured), generative AI could eventually add ~$8 trillion to the global economy each year1F 2 • VC accelerator Y Combinator is “all-in” with respect to AI, and its track record is impressive. Jared Heyman’s Rebel Fund estimated returns on every Y Combinator deal since 2005. If you had invested in each (which is practically impossible), your annual average return would have been 176% net of dilution2F 3 • AI is driving the VC ecosystem, accounting for 40%+ of new unicorns in the first half of 2024 and 60%+ of the increase in VC-backed valuations; US unicorn value grew by $162 billion so far in 2024 (Pitchbook) • AI is advancing so rapidly that performance benchmarks for reading comprehension, image classification and advanced math are becoming obsolete, with new ones replacing them (see box on GPQA benchmark) • Generative AI may not just improve productivity of call centers, coders and professional writers; a Google study projects that ~80% of jobs could see at least 10% of tasks done twice as quickly3F 4 • Elon Musk’s new xAI startup raised $6 billion at a valuation of $24 billion this year, with plans to build the world’s largest supercomputer in Tennessee to support AI training and inference efforts. Musk has been further burnishing his libertarian credentials by sharing fake AI-generated videos of VP Kamala Harris4F 5 • Then there’s the mind-altering 165-page “Situational Awareness” paper by Aschenbrenner which takes AI predictions to another level (superintelligence, 20% of US power supply, $1 trillion and full AGI by 2030)5F 6 • Rob Toews at Radical Ventures describes AlphaFold and other protein folding algorithms as the most important achievements in the history of AI (see April 2 2024 EoTM, page 3 for more details) 1 MIT Artificial Intelligence, May 2024 2 "The economic potential of generative AI: The next productivity frontier", McKinsey & Company, June 2023 3 “On the 176% annual return of a YC startup index…and why one will never exist”, Jared Heyman, July 2023. Findings: 5%- 6% of YC startups became $ 1 bn unicorns, and of those 10-12% became decacorns (10x more than unicorns). Unfortunately, it isn’t possible to invest in every Y Combinator deal since companies hand-pick investors in oversubscribed start-up rounds. In addition: those 176% average annual returns are concentrated in a subset of deals. Y Combinator’s deal failure rate is high, ~40% during its first 17 batches, and that excludes companies that are struggling or were sold at a loss 4 “The economic impact of generative AI”, Andrew McAfee, Visiting Fellow / Technology & Society, Google, April 2024 5 "Recipe for disaster: Experts issue warning after Musk shares fake AI video of Harris”, Salon.com, July 2024 6 In “Situational Awareness: the Decade Ahead” (June 2024), Leopold Aschenbrenner argues that he is among a few hundred people in the world who truly understand the scope and magnitude of the AI transformation. He explains how hundreds of millions of AI systems can automate AI research, resulting in “superintelligence” which would be the most powerful weapon that mankind has ever built…and that the democratic world would not survive China getting true AGI first 0 50 100 150 200 250 300 350 400 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 Other AI Y Combinator startups by field Count Source: Y Combinator, Ben Evans, July 2024 The Google-proof Q&A benchmark (GPQA) • Consists of 400+ multiple choice questions • Even PhD-level scholars correctly answer questions in their field only 65% of the time. The same scholars scored 34% outside their field even with unrestricted access to the internet • In 2023, most AI models scored 30%-40% on GPQA. Claude 3 (the latest Anthropic chatbot) scored ~60% • HOWEVER…this is still just an example of excellence in memorization and the problem of “contamination”, since models are first trained on data they’re tested on

• 6. EYE ON THE MARKET • MICHAEL CEMBALEST • J.P. MORGAN • September 3, 2024 2024 energy paper / US inflation/labor monitor / US Federal debt monitor / US economic monitor 6 And yet…there are a lot of questions on spending, revenues, productivity benefits, energy and more • Will OpenAI lose $5 billion this year as reported by The Information?6F 7 • Will it really cost $100 billion to train a single AI model by 2027, a figure cited by Anthropic’s CEO? 7F 8 • Will the largest tech companies be able to come up with $500 billion in annual “missing revenues” to break even vs their AI data center spend, a figure estimated by David Cahn at Sequoia? • Are hyperscalers really building enough AI compute infrastructure to power 12,000 ChatGPTs, assuming its current size of 200 mm daily active users asking 1.25 queries per day?8F 9 Barclays might be substantially underestimating the query intensity and user base of widely adopted generative AI applications, particularly if such applications entail query threads that require much more computing power…but you get the point • JPM estimates that by 2026, NVIDIA GPU sales could power the training of ~1,000 versions of GPT-5 assuming 1.7 E+26 FLOPS for each…what kind of applications will use up that much computing power? • Barclays estimates that in 2024, enough GPUs were built to generate ~$100 bn of revenues at maximum utilization rates. Actual 2024 payments by GPU users: ~$10 bn. How long will it take for this gap to close? • Why do most respondents in a NewStreet Research survey cite cost decreases of just 10% or less after generative AI adoption (see chart below)? • Why do some large enterprises whose engineers use AI coding tools still not see improvements in cycle time end-to-end software delivery? Note: only ~30% of a software developer’s time is spent coding • When do markets start thinking more about geopolitical risk given TSMC’s current role as sole supplier of advanced chips to NVIDIA?9F 10 7 "Why OpenAI Could Lose $5 Billion This Year", The Information, Amir Efrati and Aaron Holmes, July 2024 8 “In Good Company with guest Dario Amodei”, a podcast by Norges Bank Investment Management, June 2024 9 "Cloud AI Capex: FOMO or Field-Of-Dreams?", Ross Sandler, Barclays Equity Research, June 2024 10 As of August 2023, NVIDIA was sourcing all advanced AI chips from TSMC. In July 2024, NVIDIA announced that it would start diversifying and begin sourcing advanced memory chips from Samsung as well 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Sales Mkting Legal HR Supply Chain IT Software Other Cost decreases driven by generative AI adoption Percent of survey respondents Source: NewStreet Research, August 2024 <10% cost decrease 10-19% cost decrease >20% cost decrease Russian imports as % of European primary energy consumption, 2021 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% European reliance on Russian energy Global semiconductor reliance on Taiwan Global semiconductor reliance on Taiwan for advanced chips Regional reliance for critical goods Percent Source: BP Statistical Review, ROC Taiwan, Global Guardian, 2024 Sequoia: Will hyperscalers and enterprise software companies earn enough to pay for AI infrastructure? Cahn takes Nvidia’s revenues and multiplies by 2x to obtain total cost of data centers, since GPUs are roughly half the cost of data center ownership (the rest is energy, buildings, generators, etc). Cahn then multiplies by 2x again to reflect the expected 50% gross margins of major hyperscalers (i.e., Microsoft, Amazon and Google which provide cloud services via Azure, Amazon Web Services and Google Cloud Platform). Cahn assumes that Google, Apple, Microsoft and Meta will generate $10 billion each year from new AI-related revenue, and that Oracle, ByteDance, Alibaba, Tencent, X and Tesla earn another $5 billion each. The gap between new AI revenues and AI data center spend: $500 billion per year “AI’s $600 billion question”, David Cahn, Sequoia Capital, June 2024

• 7. EYE ON THE MARKET • MICHAEL CEMBALEST • J.P. MORGAN • September 3, 2024 2024 energy paper / US inflation/labor monitor / US Federal debt monitor / US economic monitor 7 • What if Daron Acemoglu at MIT is right that that the annual boost to total factor productivity from AI will be just 0.06% per year, which is quite small given ~0.6% average annual TFP growth over the last 20 years?10F 11 • Why do so many Chat GPT users access it infrequently, like me? • What about energy consumption? There’s already a dispute over Talen Energy’s partial sale of its nuclear generation to Amazon. AEP and Exelon filed a complaint with FERC citing $140 million of transmission costs now borne by PJM electricity customers. And remember: the advent of cloud computing barely changed electricity demand since data centers replaced on-premises workloads which use 2x-3x more electricity per unit of compute than the cloud. No such luck with the rise of generative AI, which needs new energy • By 2024, shouldn’t AI/machine learning do better than its recent poor performance detecting breast cancer? Human doctors identified 18 breast cancers that AI missed, while AI only identified 2 that human doctors missed along with generating of hundreds of false positives11F 12 • Will NVIDIA’s next chip rollout bring down AI server costs? The company claims that a large model will be able to be trained using 2,000 Blackwell GPUs @$35k-$40k each using 4 MW of power instead of 8,000 H100 GPUs @25k each using 15 MW (i.e., 50%+ declines in cost and energy). And what if a lot of inference tasks can be performed by cheaper CPUs rather than GPUs? 11 Acemoglu (MIT) starts by assuming that generative AI could transform ~20% of tasks in the production process, and then adjusts for estimates that only 25% of that figure could be cost-effectively automated within 10 years. He estimates AI-related labor cost savings at ~25% to arrive at an estimated total factor productivity boost of just 0.064% each year. “The Simple Macroeconomics of AI”, Daron Acemoglu, MIT, May 2024 12 “A Prospective Analysis of Screen-Detected Cancers Recalled and Not Recalled by Artificial Intelligence”, Smith et al, Journal of Breast Imaging, May 2024 0% 5% 10% 15% 20% 25% 30% 35% 40% Denmark US UK Argentina France Japan Once or twice Monthly Weekly Daily Use of ChatGPT by country Percent Source: Reuters, Ben Evans, May 2024 $0 $10,000 $20,000 $30,000 $40,000 $50,000 $60,000 $70,000 $80,000 CPU Server Google TPU v5p NVIDIA DGX H100 Other (maintenance, software etc.) Electricity Depreciation of capital cost AI server cost comparisons US$ thousands per year Source: NewStreet Research, August 2024 0 5 10 15 20 25 30 1940 1950 1960 1970 1980 1990 2000 2010 2020 2030 How realistic is a nuclear renaissance in the US? # of US nuclear plants built by year of construction start Source: Power Reactor System Database, JPMAM, December 2023 “What was I made for?” GPT and the EOTM In the September 2023 Eye on the Market (“What Was I Made For”), I explained why I’m not a frequent GPT user. I asked GPT-4 70 questions from the Eye on the Market archives on markets, economics, politics and energy. GPT4’s aggregate grade: C+. It got half the questions right but the other half wrong despite being able to web crawl, open pdfs, etc, and its mistakes were not predictable, requiring me to check every answer. Limited if any productivity benefits for me if that’s the case

• 8. EYE ON THE MARKET • MICHAEL CEMBALEST • J.P. MORGAN • September 3, 2024 2024 energy paper / US inflation/labor monitor / US Federal debt monitor / US economic monitor 8 Tracking corporate sector AI adoption so far: mostly positive but some mixed signals The Census tracks AI adoption in its Business Trends and Outlook Survey. The results are positive given steady AI adoption increases from Oct 2023 to Aug 2024, and projections through 2025 (first chart). CEO surveys also point to rising adoption: 50% of CEOs surveyed say they’re hiring for gen AI-related roles that didn't exist last year, 40% plan to increase AI spend in the next year and 70% say generative AI is their top priority12F 13. But when looking at enterprise use data from Bain, AI adoption is still concentrated in development/pilotstages; actual production use cases declined a little from October 2023 to February 2024. And according to Fed data13F 14, AI job skill penetration is still low, tracking the early days of 3D printing rather than the faster adoption rates of cloud computing and smart devices (third chart). Is the decline in tech sector employment at a time of rising payrolls a sign of AI adoption reducing tech labor intensity? Unclear; we compared changes in job listings by type against estimates of each job’s AI disruptability and found only a modest correlation. Perhaps I’m too impatient; it took 20 years for 20% of addressable retail (excluding cars and car parts) to move online as e-commerce, and AI adoption appears faster than that. It also took three years after the 2007 release of the iPhone for Uber to appear, and another five years for Uber to reach 20 mm-30 mm monthly active users. One thing is clear: AI is great for consultants. BCG now earns 20% of its revenue by helping large companies figure out what to do with generative AI14F 15. 13 IBM AI CEO Survey, May 2024; KPMG AI CEO Survey, April 2024 14 “AI and Productivity Growth”, Federal Reserve Bank of St Louis, Kalyani and Hogan, April 2024 15 "BCG says AI consulting will supply 20% of revenues this year", Financial Times, April 2024 Information Professional/Scientific/Technical Educational Services Finance and Insurance Health Care and Social Assistance Real Estate and Rental Arts/Entertainment/Recreation Administrative & Support Manufacturing Retail Trade Wholesale Trade Other Services Transportation/Warehousing Construction Accommodation & Food Services 0% 5% 10% 15% 20% 25% Next 6 months August 2024 October 2023 Census: AI adoption rates by sector and date Share of firms using AI Source: US Census Bureau, JPMAM, August 2024 Software Customer service Marketing Knowledge workers Operations IT Finance Sales HR Legal 0% 10% 20% 30% 40% 50% 60% Development/pilot Production Bain: generative AI adoption rates by stage and sector, Oct 2023 vs Feb 2024 Source: Bain, Ben Evans, July 2024 0% 20% 40% 60% 80% 100% 2007 2009 2011 2013 2015 2017 2019 2021 2023 Fed: technology penetration in job openings % of urban areas with significant job presence Source: St. Louis Fed, Kalyani & Hogan, April 2024 Smart devices Cloud computing 3D printing AI 2.7 2.8 2.9 3.0 3.1 3.2 115 117 119 121 123 125 127 129 131 133 135 2021 2022 2023 2024 US private employment Millions Millions Source: BLS, Haver, JPMAM, July 2024 Private industries exinformation services Information services

• 9. EYE ON THE MARKET • MICHAEL CEMBALEST • J.P. MORGAN • September 3, 2024 2024 energy paper / US inflation/labor monitor / US Federal debt monitor / US economic monitor 9 What are equity markets pricing in? While the surge in NVIDIA’s stock has been accompanied by rising earnings, it’s 38x P/E multiple is consistent with the rest of the technology sector which is priced at the upper end of historical ranges. Investors are bullish: short interest on tech stocks has disappeared, and P/E multiples that markets apply to AI beneficiaries are not that far below multiples applied to AI suppliers. In other words, a lot of the AI revolution appears to be priced into the equity markets. Another look: the aggregate P/E ratio on an indicative generative AI ETF15F 16 has risen to 28x this year, although it’s still below its liquidity-induced peak of 38x in 2022. At least P/E multiples are not as outlandish as in prior tech booms: current tech leaders trade at 1.8x the broad market compared to peak relative P/E multiples of 6x during the dot-com era, 2.8x during the mainframe era (1969) and 2.4x during the PC era (1983)16F 17. Good news for investors: tech sector capex and R&D is not elevated relative to cash flow, at least when compared to the pre-2000 era (fourth chart). According to Empirical Research, the ratio of Mag 7 capital spending to revenues is actually below that of the broad market, and expected to stay that way through 2026. 16 We selected an AI ETF that has a large weight to NVIDIA (~7%). Some other AI ETFs have performed poorly this year since their NVIDIA weights are lower than the NVIDIA weight in the S&P 500 17 “The tech leaders: is capex killing the CHAT”, Empirical Research, August 26 2024 0x 2x 4x 6x 8x 10x 12x 14x 0% 2% 4% 6% 8% 10% 12% 1995 2000 2005 2010 2015 2020 2025 Tech & semiconductor valuations at extremes Free cash flow yield Price to book value Source: Bloomberg, JPMAM, August 15, 2024 Tech free cash flow yields Global semi index price to book ratio Nvidia Advanced Micro Devices Super Micro Computer Arista Networks Microsoft Google Amazon Salesforce Meta Intuit Adobe 0.0x 0.5x 1.0x 1.5x 2.0x 2.5x AI suppliers vs beneficiaries P/E ratio relative to equal-weighted S&P 500, forward Source: Bloomberg, JPMAM, August 16, 2024 AI suppliers AI beneficiaries 20 23 26 29 32 35 38 2020 2021 2022 2023 2024 Valuation of indicative generative AI ETF Aggregate forward P/E ratio Source: Bloomberg, JPMAM, July 2024 0% 20% 40% 60% 80% 100% 120% 140% '90 '93 '96 '99 '02 '05 '08 '11 '14 '17 '20 '23 S&P 500 TMT investment spending as share of cash flows from operations, Percent Source: Compustat, JPMAM, June 2024. TMT = Info tech, Comm services, AMZN Capex Capex + R&D

• 10. EYE ON THE MARKET • MICHAEL CEMBALEST • J.P. MORGAN • September 3, 2024 2024 energy paper / US inflation/labor monitor / US Federal debt monitor / US economic monitor 10 COVIDIA prognosis: judgment day on AI capital spending is probably 2-3 years away • Every computing cycle entails infrastructure first, then platforms and then applications. It’s probably too soon to worry that there’s no killer generative AI app yet akin to the enterprise resource planning software of the 1990s or the search and e-commerce applications of the 2000s • But the clock is ticking: NVIDIA data center revenues as a share of market-wide capital spending are projected to reach heights seen at the peak of the mainframe era in 1969 and the dot-com boom (see chart). As a result, stakes for investors are high • The table shows how hyperscalers and other data center users now spend more on model training than on inference17F 18. In English: these companies spend more than twice as much training complex foundational models than they do on running completed production models on a day-to-day basis for their end-use customers. That will probably continue for another year or so: highly profitable hyperscalers will keep the AI capital spending machine moving on their own. Also: some AI demand will be sourced from the military which is price inelastic due to national security reasons • Within the next two years, corporate AI adoption trends shown on page 6 need to move higher (i.e., a lot more inference activity) to avoid a “metaverse” outcome for all the capital deployed. As cited earlier, there needs to be several hundred billion dollars a year of AI-related demand from the corporate sector to pay for AI infrastructure 18 Industry rule of thumb: the cost of inference is roughly the square root of the cost of training 0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 2020 2021 2022 2023 2024E 2025E 2026E 1969 2000 NVIDIA data center revenues IBM peak revenues (1969) Cisco, Lucent, Nortel peak revenues (2000) AI capital spending in context Share of market-wide capital spending Source: Empirical Research, August 2024 GPUs and Other Total Training Total $ in billions other chips AI spend AI capex and R&D Inference operating costs Microsoft $20 $20 $40 $3 $3 $6 Meta $11 $12 $23 $2 $2 $4 Google $14 $15 $29 $3 $1 $4 Amazon $8 $8 $16 $2 $1 $3 Tier 2 cloud $26 $26 $52 $8 $3 $11 Enterprises & govt $26 $26 $52 $8 $2 $10 Total $105 $105 $210 $27 $12 $39 Source: New Street Research, August 2024. Operating costs include cash operating expenses, softw are, depreciation and electricity 2024 AI data center spend 2024 data center operating costs

• 11. EYE ON THE MARKET • MICHAEL CEMBALEST • J.P. MORGAN • September 3, 2024 2024 energy paper / US inflation/labor monitor / US Federal debt monitor / US economic monitor 11 Appendix I: Current and prospective NVIDIA competitors While NVIDIA currently has 90%+ market share in advanced AI chips, there are several competitors which currently offer products that compete with NVIDIA GPUs, or plan to do so. Semiconductor companies • AMD has already developed chips that compete with NVIDIA, and has developed its own ROCm software to compete with NVIDIA’s CUDA • Intel is designing Gaudi 3 to be 1.5x faster than NVIDIA’s H100 chip and 30%-60% cheaper. Intel also is working with Google, Qualcomm and others to develop a competitor to NVIDIA’s CUDA software. Intel’s upcoming Granite Rapids chip will feature inference processing units on every core (suitable for smaller models), and AMD may follow suit • Softbank’s Arm has focused on designing and selling chip IP but is reportedly in discussions with TSMC to secure capacity to begin producing AI chips sometime in 2025 Cloud service providers • Amazon already offers Trainium and Inferentia chips to existing AWS customers alongside NVIDIA chips. AWS estimates the cost of running models on their own chips to be roughly half that of NVIDIA chips • Google already offers TPU v5p chips to GCP customers, at roughly 1/3rd the annual running cost of its NVIDIA counterpart (H100 chip). Google is NVIDIA’s closest competitor among cloud service providers • Microsoft offers the Maia 100, a cheaper alternative to the H100 for Azure customers Other AI application developers and internal users • Apple is working on an AI chip project dubbed Project ACDC for internal inference computations • Cerebras partnered with Dell to market its WSE-3 chip to compete with NVIDIA’s H100 chip. The WSE-3 chip has 50x the compute power of the H100 and can train a model 10x the size of GPT-4, but can only be purchased as part of Cerebras’s computing platform, similar to buying computing capacity from AWS • Groq, founded by a former Google TPU designer, is working on a chip optimized for ultra-fast inference for less computationally intensive models. Groq partners with Global Foundries rather than TSMC • Meta has developed its own chip, the Meta Training and Inference Accelerator (MTIA), for internal AI projects. MTIA is primarily being used to train Meta’s recommendation and ranking algorithms, but the goal is for MTIA to eventually train generative AI like Meta’s Llama • Tesla has designed its own D1 chip for self-driving car projects on their Dojo supercomputing platform • OpenAI has reportedly begun discussions with Broadcom on chip development, has hired employees from the Google TPU team and is raising capital for fab construction in partnership with Intel/TSMC/Samsung NVIDIA 92% AMD, 3% Others, 5% Market share of data center GPUs based on 2023 revenue Percent Source: IoT Analytics, 2023 OpenAI 39% Microsoft 30% AWS 8% Others 16% Market share of AI models and platforms based on 2023 revenue, Percent Source: IoT Analytics, 2023 Google, 7%

• 12. EYE ON THE MARKET • MICHAEL CEMBALEST • J.P. MORGAN • September 3, 2024 2024 energy paper / US inflation/labor monitor / US Federal debt monitor / US economic monitor 12 Appendix II: The fate of prior equity market leaders For one brief single day in June 2024, NVIDIA attained the largest market cap in the S&P 500 at $3.33 trillion, surpassing Microsoft. Market leaders of decades past typically reached an inflection point of some kind, after which their market cap declined as a share of the market (GM, IBM, Altria, Cisco, GE and Exxon). Only Microsoft and Apple continued to thrive. Notes: gray bars denote periods when each stock was the largest in the S&P 500; market cap shares computed on an end-of-quarter basis 0% 2% 4% 6% 8% 10% 12% '55 '65 '75 '85 '95 '05 '15 General Motors Source: Factset, Datastream, JPMAM, June, 2024 Share of total market cap 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% '90 '95 '00 '05 '10 '15 '20 '25 IBM Source: Factset, JPMAM, June, 2024 Share of total market cap 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% '90 '95 '00 '05 '10 '15 '20 '25 Altria Source: Factset, JPMAM, June, 2024 Share of total market cap 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% 3.5% 4.0% '90 '95 '00 '05 '10 '15 '20 '25 Cisco Source: Factset, JPMAM, June, 2024 Share of total market cap 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% 3.5% 4.0% 4.5% '90 '95 '00 '05 '10 '15 '20 '25 General Electric Source: Factset, JPMAM, June, 2024 Share of total market cap 0% 1% 2% 3% 4% 5% '90 '95 '00 '05 '10 '15 '20 '25 Exxon Source: Factset, JPMAM, June, 2024 Share of total market cap 0% 1% 2% 3% 4% 5% 6% 7% 8% '90 '95 '00 '05 '10 '15 '20 '25 Apple Source: Factset, JPMAM, June, 2024 Share of total market cap 0% 1% 2% 3% 4% 5% 6% 7% 8% '90 '95 '00 '05 '10 '15 '20 '25 Microsoft Source: Factset, JPMAM, June, 2024 Share of total market cap 0% 1% 2% 3% 4% 5% 6% 7% 1990 1995 2000 2005 2010 2015 2020 2025 Nvidia Source: Factset, JPMAM, June, 2024 Share of total market cap

• 13. EYE ON THE MARKET • MICHAEL CEMBALEST • J.P. MORGAN • September 3, 2024 2024 energy paper / US inflation/labor monitor / US Federal debt monitor / US economic monitor 13 IMPORTANT INFORMATION This material is for information purposes only. The views, opinions, estimates and strategies expressed herein constitutes Michael Cembalest’s judgment based on current market conditions and are subject to change without notice, and may differ from those expressed by other areas of JPMorgan Chase & Co. (“JPM”). This information in no way constitutes J.P. Morgan Research and should not be treated as such. Any companies referenced are shown for illustrative purposes only, and are not intended as a recommendation or endorsement by J.P. Morgan in this context. GENERAL RISKS & CONSIDERATIONS Any views, strategies or products discussed in this material may not be appropriate for all individuals and are subject to risks. Investors may get back less than they invested, and past performance is not a reliable indicator of future results. Asset allocation/diversification does not guarantee a profit or protect against loss. Nothing in this material should be relied upon in isolation for the purpose of making an investment decision. NON-RELIANCE Certain information contained in this material is believed to be reliable; however, JPM does not represent or warrant its accuracy, reliability or completeness, or accept any liability for any loss or damage (whether direct or indirect) arising out of the use of all or any part of this material. No representation or warranty should be made with regard to any computations, graphs, tables, diagrams or commentary in this material, which are provided for illustration/ reference purposes only. Any projected results and risks are based solely on hypothetical examples cited, and actual results and risks will vary depending on specific circumstances. Forward-looking statements should not be considered as guarantees or predictions of future events. Nothing in this document shall be construed as giving rise to any duty of care owed to, or advisory relationship with, you or any third party. Nothing in this document shall be regarded as an offer, solicitation, recommendation or advice (whether financial, accounting, legal, tax or other) given by J.P. Morgan and/or its officers or employees,. J.P. Morgan and its affiliates and employees do not provide tax, legal or accounting advice. You should consult your own tax, legal and accounting advisors before engaging in any financial transactions. For J.P. Morgan Asset Management Clients: J.P. Morgan Asset Management is the brand for the asset management business of JPMorgan Chase & Co. and its affiliates worldwide. To the extent permitted by applicable law, we may record telephone calls and monitor electronic communications to comply with our legal and regulatory obligations and internal policies. Personal data will be collected, stored and processed by J.P. Morgan Asset Management in accordance with our privacy policies at https://am.jpmorgan.com/global/privacy. ACCESSIBILITY For U.S. only: If you are a person with a disability and need additional support in viewing the material, please call us at 1-800-343-1113 for assistance. This communication is issued by the following entities: In the United States, by J.P. Morgan Investment Management Inc. or J.P. Morgan Alternative Asset Management, Inc., both regulated by the Securities and Exchange Commission; in Latin America, for intended recipients’ use only, by local J.P. Morgan entities, as the case may be.; in Canada, for institutional clients’ use only, by JPMorgan Asset Management (Canada) Inc., which is a registered Portfolio Manager and Exempt Market Dealer in all Canadian provinces and territories except the Yukon and is also registered as an Investment Fund Manager in British Columbia, Ontario, Quebec and Newfoundland and Labrador. In the United Kingdom, by JPMorgan Asset Management (UK) Limited, which is authorized and regulated by the Financial Conduct Authority; in other European jurisdictions, by JPMorgan Asset Management (Europe) S.à r.l. In Asia Pacific (“APAC”), by the following issuing entities and in the respective jurisdictions in which they are primarily regulated: JPMorgan Asset Management (Asia Pacific) Limited, or JPMorgan Funds (Asia) Limited, or JPMorgan Asset Management Real Assets (Asia) Limited, each of which is regulated by the Securities and Futures Commission of Hong Kong; JPMorgan Asset Management (Singapore) Limited (Co. Reg. No. 197601586K), which this advertisement or publication has not been reviewed by the Monetary Authority of Singapore; JPMorgan Asset Management (Taiwan) Limited; JPMorgan Asset Management (Japan) Limited, which is a member of the Investment Trusts Association, Japan, the Japan Investment Advisers Association, Type II Financial Instruments Firms Association and the Japan Securities Dealers Association and is regulated by the Financial Services Agency (registration number “Kanto Local Finance Bureau (Financial Instruments Firm) No. 330”); in Australia, to wholesale clients only as defined in section 761A and 761G of the Corporations Act 2001 (Commonwealth), by JPMorgan Asset Management (Australia) Limited (ABN 55143832080) (AFSL 376919). For all other markets in APAC, to intended recipients only. For J.P. Morgan Private Bank Clients: ACCESSIBILITY J.P. Morgan is committed to making our products and services accessible to meet the financial services needs of all our clients. Please direct any accessibility issues to the Private Bank Client Service Center at 1-866-265-1727 LEGAL ENTITY, BRAND & REGULATORY INFORMATION In the United States, JPMorgan Chase Bank, N.A. and its affiliates (collectively “JPMCB”) offer investment products, which may include bank managed investment accounts and custody, as part of its trust and fiduciary services. Other investment products and services, such as brokerage and advisory accounts, are offered through J.P. Morgan Securities LLC (“JPMS”), a member of FINRA and SIPC. JPMCB and JPMS are affiliated companies under the common control of JPM. In Germany, this material is issued by J.P. Morgan SE, with its registered office at Taunustor 1 (TaunusTurm), 60310 Frankfurt am Main, Germany, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB). In Luxembourg, this material is issued by J.P. Morgan SE – Luxembourg Branch, with registered office at European Bank and Business Centre, 6 route de Treves, L-2633, Senningerberg, Luxembourg, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB); J.P. Morgan SE – Luxembourg Branch is also supervised by the Commission de Surveillance du Secteur Financier (CSSF); registered under R.C.S Luxembourg B255938. In the United Kingdom, this material is issued by J.P. Morgan SE – London Branch, registered office at 25 Bank Street, Canary Wharf, London E14 5JP, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB); J.P. Morgan SE – London Branch is also supervised by the Financial Conduct Authority and Prudential Regulation Authority. In Spain, this material is distributed by J.P. Morgan SE, Sucursal en España, with registered office at Paseo de la Castellana, 31, 28046 Madrid, Spain, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB); J.P. Morgan SE, Sucursal en España is also supervised by the Spanish Securities Market Commission (CNMV); registered with Bank of Spain as a branch of J.P. Morgan SE under code 1567. In Italy, this material is distributed by J.P. Morgan SE – Milan Branch, with its registered office at Via Cordusio, n.3, Milan 20123, Italy, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin,

• 14. EYE ON THE MARKET • MICHAEL CEMBALEST • J.P. MORGAN • September 3, 2024 2024 energy paper / US inflation/labor monitor / US Federal debt monitor / US economic monitor 14 the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB); J.P. Morgan SE – Milan Branch is also supervised by Bank of Italy and the Commissione Nazionale per le Società e la Borsa (CONSOB); registered with Bank of Italy as a branch of J.P. Morgan SE under code 8076; Milan Chamber of Commerce Registered Number: REA MI 2536325. In the Netherlands, this material is distributed by J.P. Morgan SE – Amsterdam Branch, with registered office at World Trade Centre, Tower B, Strawinskylaan 1135, 1077 XX, Amsterdam, The Netherlands, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB); J.P. Morgan SE – Amsterdam Branch is also supervised by De Nederlandsche Bank (DNB) and the Autoriteit Financiële Markten (AFM) in the Netherlands. Registered with the Kamer van Koophandel as a branch of J.P. Morgan SE under registration number 72610220. In Denmark, this material is distributed by J.P. Morgan SE – Copenhagen Branch, filial af J.P. Morgan SE, Tyskland, with registered office at Kalvebod Brygge 39-41, 1560 København V, Denmark, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB); J.P. Morgan SE – Copenhagen Branch, filial af J.P. Morgan SE, Tyskland is also supervised by Finanstilsynet (Danish FSA) and is registered with Finanstilsynet as a branch of J.P. Morgan SE under code 29010. In Sweden, this material is distributed by J.P. Morgan SE – Stockholm Bankfilial, with registered office at Hamngatan 15, Stockholm, 11147, Sweden, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB); J.P. Morgan SE – Stockholm Bankfilial is also supervised by Finansinspektionen (Swedish FSA); registered with Finansinspektionen as a branch of J.P. Morgan SE. In Belgium, this material is distributed by J.P. Morgan SE – Brussels Branch with registered office at 35 Boulevard du Régent, 1000, Brussels, Belgium, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB); J.P. Morgan SE Brussels Branch is also supervised by the National Bank of Belgium (NBB) and the Financial Services and Markets Authority (FSMA) in Belgium; registered with the NBB under registration number 0715.622.844. In Greece, this material is distributed by J.P. Morgan SE – Athens Branch, with its registered office at 3 Haritos Street, Athens, 10675, Greece, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB); J.P. Morgan SE – Athens Branch is also supervised by Bank of Greece; registered with Bank of Greece as a branch of J.P. Morgan SE under code 124; Athens Chamber of Commerce Registered Number 158683760001; VAT Number 99676577. In France, this material is distributed by J.P. Morgan SE – Paris Branch, with its registered office at 14, Place Vendôme 75001 Paris, France, authorized by the Bundesanstaltfür Finanzdienstleistungsaufsicht(BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB) under code 842 422 972; J.P. Morgan SE – Paris Branch is also supervised by the French banking authorities the Autorité de Contrôle Prudentiel et de Résolution (ACPR) and the Autorité des Marchés Financiers (AMF). In Switzerland, this material is distributed by J.P. Morgan (Suisse) SA, with registered address at rue du Rhône, 35, 1204, Geneva, Switzerland, which is authorised and supervised by the Swiss Financial Market Supervisory Authority (FINMA) as a bank and a securities dealer in Switzerland. In Hong Kong, this material is distributed by JPMCB, Hong Kong branch. JPMCB, Hong Kong branch is regulated by the Hong Kong Monetary Authority and the Securities and Futures Commission of Hong Kong. In Hong Ko.ng, we will cease to use your personal data for our marketing purposes without charge if you so request. In Singapore, this material is distributed by JPMCB, Singapore branch. JPMCB, Singapore branch is regulated by the Monetary Authority of Singapore. Dealing and advisory services and discretionary investment management services are provided to you by JPMCB, Hong Kong/Singapore branch (as notified to you). Banking and custody services are provided to you by JPMCB Singapore Branch. The contents of this document have not been reviewed by any regulatory authority in Hong Kong, Singapore or any other jurisdictions. You are advised to exercise caution in relation to this document. If you are in any doubt about any of the contents of this document, you should obtain independent professional advice. For materials which constitute product advertisement under the Securities and Futures Act and the Financial Advisers Act, this advertisement has not been reviewed by the Monetary Authority of Singapore. JPMorgan Chase Bank, N.A., a national banking association chartered under the laws of the United States, and as a body corporate, its shareholder’s liability is limited. With respect to countries in Latin America, the distribution of this material may be restricted in certain jurisdictions. Issued in Australia by JPMorgan Chase Bank, N.A. (ABN 43 074 112 011/AFS Licence No: 238367) and J.P. Morgan Securities LLC (ARBN 109293610). References to “J.P. Morgan” are to JPM, its subsidiaries and affiliates worldwide. “J.P. Morgan Private Bank” is the brand name for the private banking business conducted by JPM. This material is intended for your personal use and should not be circulated to or used by any other person, or duplicated for non-personal use, without our permission. If you have any questions or no longer wish to receive these communications, please contact your J.P. Morgan team. © 2024 JPMorgan Chase & Co. All rights reserved.