AI has become the tech industry’s core capital project, and 2025 is the year the spend shows up everywhere from grid loads to OEM backlogs. Microsoft is guiding to record quarterly capex of roughly $30 billion as AI demand compounds in Azure; Meta lifted its full-year 2025 capex outlook to $66–$72 billion to accelerate compute and data-center build-outs; and Amazon is pacing toward ~$100 billion of total capex this year as it scales AWS and enterprise AI contracts.
Together with peers, hyperscaler investment tied to AI infrastructure is widely expected to run into the low-$300-billions in 2025. The common thread: capacity first, monetization second.
Throughput is validating the thesis. NVIDIA reported $46.7 billion in revenue last quarter, with data-center sales around $41 billion as the Blackwell cycle ramps — evidence that training and serve capacity remains the tightest bottleneck in the stack. On the demand side, AI is moving from pilots to production: IDC now projects AI solution spending at roughly $307 billion in 2025, expanding toward ~$632 billion by 2028, while a broader Gartner lens pegs worldwide AI spend near $1.5 trillion in 2025. Different scopes, same direction: the wallet share is real and rising.
Unit economics are tilting decisively toward inference. Stanford’s 2025 AI Index reports that inference costs for GPT-3.5-class performance dropped over 280× between late 2022 and late 2024, a collapse driven by smaller, more efficient model architectures and optimized compute pipelines.
The same report notes hardware costs falling about 30% per year and energy efficiency improving roughly 40% annually, reinforcing the shift toward “serve-side” AI that runs cheaply and ubiquitously inside software stacks. The economic effect is simple: as the marginal cost of inference plummets, usage scales exponentially, and revenue increasingly accrues to firms pricing by tokens, requests, or latency guarantees rather than static seat licenses.
Two real-world constraints now shape the cycle’s leaders. First is power. The IEA projects global data-center electricity consumption more than doubling by 2030 to ~945 TWh, with AI the largest incremental driver; in the U.S., the EIA already sees record power use in 2025–26 as data centers, AI, and electrification stack. Geography with fast interconnects, substation lead-times, and cheap megawatts will attract outsized build-outs; everyone else will ration. Second is memory bandwidth. HBM supply is tight and strategically controlled: SK hynix led Q2 2025 share (~62%), Micron ~21%, Samsung ~17%, and industry expectations point to a ~$100 billion HBM market by 2030 as models demand wider pipes and better packaging yields. These choke points tend to accrue pricing power and margin to whoever can secure capacity and deliver yield.
Netting it out: the fastest-growing corner of “AI” is not a product category but a flywheel, capex → capacity → tokens served → software attach → more capex. 2025’s tell is that supply-side constraints (power, HBM, advanced packaging) and serve-side economics (inference cost curves, latency guarantees) now explain more of the variance in growth than model-benchmark leapfrogging alone.
For investors, that argues for selection grounded in constraint-aware exposure (power, memory bandwidth, efficient networking) and in real adoption telemetry (usage-linked revenue, contract wins, workload migration), not slideware. The secular uptrend is intact; the edge lies in underwriting where tokens and watts convert to dollars soonest.
Source: unsplash
Methodology
For our list of the fastest-growing AI stocks to invest in now, we picked pure-play AI stocks and/or AI-adjacent stocks that had the highest year-over-year revenue growth and ranked them as such. We sourced the growth numbers from stockanalysis.com. We have also mentioned the number of hedge funds holding stake in these stocks as of Q2, 2025.
Some of these stocks may have negative EPS, but that’s common in high-growth tech where scale comes first. These firms burn cash on R&D and expansion, betting that today’s losses fuel tomorrow’s dominance and operating leverage.
Why are we interested in the stocks that hedge funds pile into? The reason is simple: our research has shown that we can outperform the market by imitating the top stock picks of the best hedge funds. Our quarterly newsletter’s strategy selects 14 small-cap and large-cap stocks every quarter and has returned 427.7% since May 2014, beating its benchmark by 264 percentage points (see more details here).
10. Confluent, Inc. (NASDAQ:CFLT)
Year-over-Year Revenue Growth: 23%
Number of Hedge Fund Holders: 50
Confluent, Inc. (NASDAQ:CFLT) is one of the fastest-growing AI stocks to invest in now. On October 9, Citi’s Tyler Radke maintained Neutral on Confluent and raised the 12-month price target to $25 (from $20).
On September 30, Wells Fargo initiated coverage at Overweight with a $24 target under analyst Ryan Macwilliams. In press summaries of the initiation, Wells Fargo characterized Confluent as an “underappreciated AI beneficiary” and argued that DIY Kafka-based builds are often overstated as a cheaper alternative at scale.
Confluent, Inc. (NASDAQ:CFLT) was founded in 2014 by the original creators of Apache Kafka: Jay Kreps, Jun Rao, and Neha Narkhede, and provides a managed data-streaming platform (Confluent Cloud/Platform) widely used to move and process real-time event data that underpins analytics and AI workloads.
9. SentinelOne, Inc. (NYSE:S)
Year-over-Year Revenue Growth: 25%
Number of Hedge Fund Holders: 45
SentinelOne, Inc. (NYSE:S) is one of the fastest-growing AI stocks to invest in now. In recent weeks, three major analysts issued fresh ratings on the company, each offering a different shade of market conviction. On October 3, 2025, Shaul Eyal of TD Cowen maintained a Buy rating with a $24.00 price target, pointing to SentinelOne’s expanding growth potential and strategic positioning in AI‑driven cybersecurity.
Earlier, on September 16, Fatima Boolani of Citi reiterated a Hold rating, raising her price target from $19 to $20. Her cautious tone suggests some concern around competitive dynamics and near-term volatility.
More decisively bullish was Trevor Walsh of JMP Securities, who on September 9, 2025, maintained a Market Outperform rating with a $29.00 price target. His rating followed SentinelOne’s strong Q2 FY2026 earnings, where revenue hit $242.2 million and Annual Recurring Revenue crossed the $1 billion mark.
Founded in 2013, SentinelOne, Inc. (NYSE:S) uses behavioral AI to autonomously detect and neutralize threats across cloud, identity, and endpoint layers. It continues to expand its security stack through focused acquisitions and platform development.
