US stocks are back at record highs, with the S&P 500 clearing 6,500 as AI heavyweights extend their lead. Into that momentum, Evercore ISI says the index can climb another 20% by end-2026 on sustained artificial intelligence enthusiasm and spending. The bet lands as Nvidia, Palantir, Broadcom and AMD keep outrunning the market, even while the AI bellwether’s stock slipped after blockbuster results. The question is not whether AI is big. It is whether earnings can rise fast enough across the index to justify another leg up.
A 10% gain this year and fresh records give Evercore’s call immediate relevance. The market already priced in an AI buildout that has turned mega-cap balance sheets into infrastructure funds. The index is more concentrated, richer, and more dependent on a narrow group of enablers than at any time since the peak of the pre-cloud era. Supporters argue that this time the capacity being built has identifiable buyers and clear use cases. Skeptics counter that the multiples assume diffusion of benefits to the rest of the S&P that remains largely unproven. That’s the fault line Evercore is straddling.
This is a capex cycle call. Hyperscalers Microsoft, Meta, Amazon and Alphabet are driving the order book for Nvidia’s data center chips, together accounting for roughly 40% of revenue at the AI leader. Those firms have signaled multi-year outlays to build and train models, stand up inference capacity, and push AI into search, ads, e-commerce, productivity suites, and social feeds. If that spending persists, the supply chain — from Nvidia (NVDA) and AMD (AMD) to Broadcom (AVGO) and high-end networking — continues to see throughput. The bet is that the spending runs long enough, and the monetization is strong enough, to lift earnings not just for the builders, but for the adopters across software, industrials, health care and finance.
For the S&P 500 to tack on 20% without a rate tailwind, either multiples expand from already rich levels or profits rise at a clip that outpaces history. A durable AI-driven rally likely needs the latter. That means measurable productivity improvements and new revenue lines at the non-mega-cap majority of the index. Margin expansion from automation, better customer acquisition via AI-driven targeting, and faster product cycles are the promised bridges. If those show up in quarterly results and guidance, the Evercore path opens. If not, the rally risks devolving into multiple inflation atop a narrow earnings base — a setup prone to sharp reversals.
Nvidia’s latest quarter fit the bull script on paper: data center and AI chip revenue jumped 56% year over year to $41.1 billion. Yet the stock fell about 3% after the report. That reaction captured the market’s tension. Delivery remains exceptional, but the bar is moving higher as investors probe the durability of hyperscaler orders, the cadence of new GPU cycles, and how quickly inference workloads translate to volume rather than one-off training spikes. As the AI bellwether, Nvidia’s post-earnings selloff rippled through peers and underscored how sensitive the trade has become to any hint of saturation or normalization in pricing and supply.
A recent MIT analysis found about 95% of companies haven’t yet realized meaningful returns on their AI investments. That does not undermine the long-term case, but it does highlight a timing mismatch. CFOs can fund pilots for a year or two; by year three, ROI pressure gets real. If enterprise adoption remains stuck in proof-of-concept mode, the demand curve could flatten faster than the capex plans assume. That gap is where the Evercore call could falter. The next phase needs examples of cost savings and revenue uplift that shareholders can see in gross margin and operating income. Without that, AI becomes a cost line with an uncertain payback.
If AI optionality is the defining multiple driver, Tesla (TSLA) sits in the crosshairs. It is a carmaker valued in part as an autonomy and robotics platform, with software upgrade economics and potential robotaxi or humanoid robotics revenue that bulls view as underappreciated. Any credible step-change on full self-driving or AI-enabled services reignites the narrative and drags the entire AI complex with it. The flip side is equally sharp: delays, regulatory setbacks, or safety incidents in autonomy would compress the AI premium embedded in Tesla and ripple through sentiment for AI-adjacent bets. That asymmetry is a feature of this tape and a risk to any smooth glide path to Evercore’s target.
Macro still matters. If inflation proves sticky and pushes yields higher, long-duration growth assets will struggle to hold expanded multiples. Regulatory scrutiny on data usage and AI-generated content could slow rollouts or add cost. GPU supply is expanding, and competition is intensifying, which could pressure pricing and margins across the stack. If the hyperscalers pivot from build to optimize faster than expected, suppliers face an air pocket. Finally, breadth remains thin; if leadership narrows further, index-level gains rely on an even smaller group of names, increasing fragility.
Into year-end and early 2026, the markers are clear. Watch capex commentary and AI revenue disclosures from Microsoft (MSFT), Amazon (AMZN), Alphabet (GOOGL) and Meta (META). Track Nvidia’s order visibility and any signals on inference demand versus training. Monitor Broadcom’s custom silicon wins and AMD’s MI300 traction as barometers of a multi-vendor landscape. Outside tech, look for concrete AI-driven efficiency metrics — cycle-time reductions, headcount productivity, customer support deflection rates — and whether they flow through to earnings. Keep an eye on market breadth indicators and the gap between mega-cap and equal-weight S&P performance. If adoption accelerates and profits broaden, Evercore’s 20% roadmap looks less like euphoria and more like math. If not, the index has already cashed a sizable AI check.
