
Proprietary AI analyst is now available
- Insights
- Public AI Valuation Multiples — June 2026
Public AI Valuation Multiples — June 2026
Public AI valuations as of June 2026
As of June 2026, the AI complex does not trade as one block. The 9 hyperscalers and “Big AI” names at the top of the stack change hands at a median 11.1x forward revenue; the semiconductor supply chain sits at 10.5x, neoclouds and data centers at 11.8x, and networking at 12.0x - while the server and hardware-assembly layer trades for just 0.9x, a fraction of the rest of the stack.
On forward EBITDA the order reshuffles. The supply chain (31.4x) and networking (30.7x) carry the richest multiples - the scarcity premium for leading-edge fabrication and high-speed interconnect - while memory (11.2x) and assembly (12.0x) anchor the bottom, the layers where pricing is commoditized and swings with the cycle.
Growth only partly explains the spread. Neoclouds are the fastest-growing group in the set (median 103% revenue growth) and are paid for it. Memory grows at a healthy 46% yet trades at barely half the neocloud revenue multiple (5.8x versus 11.8x) - the market pricing a cyclical peak, not a secular one. The sections below take each segment in turn.
Hyperscalers and "Big AI" valuation multiples in June 2026
AI conglomerates and hyperscalers are the largest and most diversified beneficiaries of the AI build-out. This space includes chip designers whose compute trains frontier models, and the hyperscalers that monetize AI through cloud, advertising, and software. Their AI exposure sits on top of enormous, cash-generative core franchises.
Because incumbency and balance-sheet strength anchor these names, public markets value them at premium (that is while scrutinizing how durable AI-driven growth really is and how much capital is being committed to compute).
NVIDIA here obviously leads a pack, valued at $5.0T EV, or 11.1x NTM revenue, followed by Alphabet ($4.5T, 8.4x) and Microsoft ($2.8T, 7.3x).
# | |||||||
|---|---|---|---|---|---|---|---|
1 | $5.0T | 11.1x | 16.5x | 61% | 148% | $5.1M | |
2 | $4.5T | 8.4x | 17.8x | 20% | 67% | $2.1M | |
3 | $2.8T | 7.3x | 11.7x | 17% | 78% | $1.2M | |
4 | $2.7T | 3.1x | 11.6x | 14% | 40% | $455K | |
5 | $2.0T | 13.5x | 19.7x | 65% | 133% | $1.9M | |
6 | $868B | 13.9x | 46.4x | 49% | 70% | $1.1M | |
7 | $654B | 7.2x | 12.8x | 33% | 83% | $354K | |
8 | $273B | 20.3x | 50.9x | 42% | 80% | $1.2M | |
9 | $50.7B | 28.9x | 127.6x | – | 43% | $1.3M | |
| Median | 11.1x | 17.8x | 38% | 78% | $1.2M |
Servers and AI infrastructure valuation multiples in June 2026
AI infrastructure companies build and assemble the physical systems AI runs on, so servers, racks, and the manufacturing services behind them.
Hardware assembly is competitive and capex-intensive, that translates to thinner margins than chip designers or software platforms.
They are generally valued bit lower than the broader AI, with valuations sensitive to order backlogs, customer concentration, and the demand - supply risk.
Dell is the highest valued public AI infra company at $284B EV (1.6x NTM revenue), ahead of Flex ($56.0B, 1.6x) and Celestica ($43.3B, 1.9x). Multiples across the group are compressed, consistent with the contract-manufacturing economics for the sector.
# | |||||||
|---|---|---|---|---|---|---|---|
1 | $284B | 1.6x | 13.9x | 32% | 58% | $1.1M | |
2 | $56.0B | 1.6x | 17.3x | 22% | 25% | $189K | |
3 | $43.3B | 1.9x | 20.7x | 46% | 59% | $461K | |
4 | $41.8B | 1.0x | 12.9x | 21% | 27% | $298M | |
5 | $37.8B | 0.4x | 7.2x | 11% | 19% | $959K | |
6 | $28.7B | 0.5x | 6.4x | 2% | 12% | $1.0M | |
7 | $27.3B | 0.5x | 9.4x | 30% | 53% | $3.5M | |
8 | $13.6B | 0.9x | 11.1x | 26% | 60% | $189K | |
| Median | 0.9x | 12.0x | 24% | 40% | $982K |
Neoclouds and data center valuation multiples in June 2026
These companies own or operate the compute capacity and physical real estate AI workloads depend on, like specialized GPU clouds, colocation and interconnection. The economics are very capex-heavy and financed by long-lived assets.
Investors weigh utilization, contract duration, and the cost of capital as much as headline growth. Established data-center operators trade on stable, recurring cash flows, while newer GPU-cloud operators are valued on growth expectations that carry both higher upside and greater execution and financing risk.
Equinix leads with EV of $128B (11.9x NTM revenue), followed by CoreWeave ($97.2B, 5.3x) and Digital Realty Trust ($83.6B, 11.8x).
Big valuation spread between established colocation operators and newer GPU-cloud entrants (the latter carrying the more premium multiples.)
# | |||||||
|---|---|---|---|---|---|---|---|
1 | $128B | 11.9x | 23.3x | 11% | 62% | $672K | |
2 | $97.2B | 5.3x | 8.7x | 113% | 205% | – | |
3 | $83.6B | 11.8x | 21.5x | 11% | 65% | $1.4M | |
4 | $73.0B | 10.1x | 19.5x | 278% | 599% | $386K | |
5 | $23.2B | 8.2x | 11.6x | 290% | 259% | $1.9M | |
6 | $18.6B | 13.2x | 33.4x | 40% | 65% | $617K | |
7 | $14.4B | 16.3x | 42.1x | 103% | 149% | $703K | |
| Median | 11.8x | 21.5x | 103% | 149% | $688K |
Semiconductor supply chain valuation multiples in June 2026
AI picks-and-shovels layer - foundries, equipment makers, materials, packaging, and test that every advanced chip must pass through. These businesses tend to have deep technical moats and extremely high barriers to entry.
That supports durable margins, but the group is cyclical and exposed to the timing of fab investment. The market values critical, hard-to-replicate tooling and leading-edge manufacturing at a premium, while more commoditized or trailing-edge participants trade at a discount.
TSMC dominates the space at $1.9T EV (10.2x NTM revenue), with ASML ($728B, 14.6x) and Intel ($686B, 11.1x) close behind. Biggest premium goes to the "irreplaceable" names owning leading-edge manufacturing.
# | |||||||
|---|---|---|---|---|---|---|---|
1 | $1.9T | 10.2x | 14.0x | 32% | 110% | $1.4M | |
2 | $728B | 14.6x | 35.5x | 19% | 59% | $887K | |
3 | $686B | 11.1x | 31.2x | 11% | 44% | $621K | |
4 | $489B | 12.5x | 35.0x | 24% | 54% | $777K | |
5 | $486B | 15.9x | 40.1x | 32% | 68% | $970K | |
6 | $340B | 20.0x | 42.2x | 26% | 66% | $800K | |
7 | $212B | 10.2x | 31.7x | 26% | 53% | $774K | |
8 | $89.5B | 3.3x | 14.0x | 21% | 44% | $185M | |
9 | $68.4B | 13.7x | 38.7x | 31% | 76% | $483K | |
10 | $55.2B | 5.9x | 13.2x | 15% | 60% | $399K | |
11 | $46.4B | 6.1x | 17.7x | 9% | 40% | $624K | |
12 | $31.2B | 6.1x | 22.4x | 19% | 49% | $382K | |
13 | $31.0B | 14.1x | 36.3x | 28% | 58% | $292K | |
14 | $30.5B | 8.4x | 28.8x | 10% | 36% | $415K | |
15 | $22.1B | 2.8x | 14.8x | 12% | 32% | $218K | |
16 | $16.0B | 10.8x | 31.7x | 28% | 93% | $831K | |
17 | $14.9B | 6.2x | 26.9x | 21% | 46% | $138K | |
18 | $8.8B | 13.9x | 44.5x | 19% | 45% | $2.9M | |
| Median | 10.5x | 31.4x | 21% | 54% | $699K |
Memory and data storage valuation multiples in June 2026
Memory and storage supply the capacity AI’s data-hungry workloads consume, spanning DRAM, NAND flash, hard drives, and storage systems. Historically this has been the most cyclical corner of semiconductors, with commodity-like pricing that swings on supply and demand.
As a result, the market typically applies lower (more volatile) valuation multiples here than to differentiated logic or software.
Samsung is the largest by enterprise value at $1.4T (2.8x NTM revenue), ahead of Micron ($1.3T, 7.0x) and SK Hynix ($1.2T, 4.7x).
# | |||||||
|---|---|---|---|---|---|---|---|
1 | $1.4T | 2.8x | 4.7x | 50% | 164% | $1.8M | |
2 | $1.3T | 7.0x | 8.3x | 87% | 209% | $705K | |
3 | $1.2T | 4.7x | 5.7x | 89% | 328% | $2.1M | |
4 | $334B | 5.4x | 6.6x | 154% | 324% | $706K | |
5 | $320B | 7.2x | 9.9x | 128% | 208% | $669K | |
6 | $256B | 14.5x | 28.9x | 38% | 83% | $238K | |
7 | $245B | 14.9x | 30.4x | 38% | 79% | $303K | |
8 | $30.4B | 4.0x | 12.5x | 8% | 38% | $592K | |
9 | $23.5B | 5.0x | 21.4x | 19% | 44% | $610K | |
10 | $10.7B | 6.2x | 25.2x | 41% | 107% | $487K | |
| Median | 5.8x | 11.2x | 46% | 136% | $640K |
Networking valuation multiples in June 2026
These companies supply the optical and high-speed connectivity - transceivers, lasers, cabling, fiber, and interconnect. Public markets reward proprietary tech and rising content per system with higher multiples, while more commoditized component suppliers see valuations that track the broader hardware cycle.
Zhongji Innolight tops the group at $224B EV (12.2x NTM revenue), followed by Amphenol ($216B, 6.1x) and Corning ($176B, 8.6x).
# | |||||||
|---|---|---|---|---|---|---|---|
1 | $224B | 12.2x | 30.5x | 91% | 192% | $663K | |
2 | $216B | 6.1x | 18.7x | 28% | 77% | $136K | |
3 | $176B | 8.6x | 29.6x | 16% | 50% | $244K | |
4 | $119B | 12.3x | 26.7x | 75% | 155% | $705K | |
5 | $77.2B | 8.1x | 31.0x | 36% | 55% | $192K | |
6 | $70.3B | 37.9x | 99.3x | 58% | 120% | $1.1M | |
7 | $66.3B | 11.9x | 28.1x | 89% | 126% | $156K | |
8 | $49.3B | 19.2x | 36.2x | 73% | 148% | $2.1M | |
9 | $29.6B | 19.8x | 55.9x | 27% | 64% | $484K | |
10 | $19.6B | 3.5x | 27.7x | 23% | 40% | $208K | |
11 | $12.8B | 6.9x | 52.0x | 156% | 139% | $138K | |
12 | $5.5B | 30.7x | 146.5x | 57% | 77% | $82K | |
| Median | 12.0x | 30.7x | 57% | 98% | $226K |
Data and methodology
Underlying data
Public markets data is powered by FactSet (consensus analyst estimates), and Morningstar (historical data). Data points are calendarized to December where relevant: retrieved data on financial year ends (e.g. FY, FY+1 etc.) are mapped to calendar years (2025A, 2026E etc.) before the appropriate month weights are then applied to prior/future fundamentals.
Private transaction data is multi-sourced, aggregated from harvesting public information, 3rd party APIs, and data engineering. All data is verified and provided with an extensive manual process. If data permits, we apply our own logic to get to the EV. For example, for a large M&A deal with available information on the target's net debt, we might adjust a valuation to fully reflect an accurate EV. In all other cases, we take the reported valuation as the numerator. Financials: we source LTM revenue and LTM EBITDA data from company filings, press releases, or other verified sources. If LTM data is unavailable, we take the 'next best-fit' period (run-rate or calendar year), provided it makes sense in a given case. For example, if a deal closed in November 2025, we might take full-year 2025 revenue as a revenue benchmark.
Any raw figures are harmonised to USD for comparison purposes.
This page is built on data available at Multiples. Sign up for a free trial to see the full dataset.
Start Your
Free Trial Today
Try Multiples for free for 3 days. Got questions or need a demo? Schedule a call with us below.













