Loading analysis...
AKAM Revalues AI Edge Inference Through Major Anthropic Deal
Show original post
Alright regards, strap in. I know what you're thinking. "Akamai? My dad's router has more sex appeal than this internet grandpa company." Bear with me.
THE TLDR FOR THE ILLITERATE
This 25-year-old CDN dinosaur just landed a $1.8 BILLION deal with Anthropic, ripped 27% in one day - the biggest single-day move in 22 YEARS - and is still trading at 1/10th the valuation of Cloudflare. Either Cloudflare is insane or Akamai is criminally underpriced. Probably both. Welcome to the market.
BUT FIRST - THE MACRO SETUP
Before we get into Akamai, let me tell you what's happening in the background that makes this the right trade right now.
Semiconductors and memory are absolutely ripping. DRAM prices surged 246% year-over-year in 2025. HBM (the memory that sits right next to your GPU and makes it not useless) is in shortage so severe that analysts are calling it the strongest memory cycle in semiconductor history. Memory spending is projected to jump from $216 billion to $633 billion in 2026 alone. The whole chip industry is on track to hit $1 TRILLION in annual sales.
So what does this mean for AI? Running AI just got really, really expensive and it's not getting cheaper anytime soon.
Here's what's actually happening at Anthropic right now, and this is where the Akamai story starts connecting. Claude users are hitting walls. Rate limits are tighter than they've ever been. Pro plan users are burning through their allowance in hours. Claude Code was eating so much compute that Anthropic briefly tried to yeet it off the $20 plan entirely - users revolted, Anthropic walked it back in under 24 hours, and their head of growth had to publicly apologize.
Why is this happening? Every time Anthropic ships a better model it's also a more compute-hungry model. Opus 4.7 just dropped a new tokenizer that uses up to 35% MORE tokens for the same text meaning your bill quietly went up 35% even though the price per token didn't change. Fast Mode runs at a 6x price premium. US-only data residency costs 10% extra on top of that. The pricing is getting more complex and more expensive at every layer while users are getting more rate limited, not less.
The $200/month Max plan can generate what would cost $5,000 at retail API prices for a power user. Anthropic is basically subsidizing inference at scale because their owned infrastructure can't keep up with demand and new data center capacity won't come online until late 2026 at the earliest.
So here's where Anthropic is stuck: exploding demand, users screaming about rate limits, and GPU/memory costs going vertical against them.
They need cheaper inference. Not eventually. Yesterday.
WHAT DOES AKAMAI EVEN DO
Three businesses:
- Security (~55% of revenue, growing 11% YoY) - Bouncer for the entire internet. WAF, DDoS protection, zero-trust, API security. Their API Security and Guardicore products grew 36% last quarter. THIRTY. SIX. PERCENT.
- Cloud Infrastructure / CIS (~9% of revenue, growing 40% YoY) - The money printer. Where Anthropic just planted $1.8B. Four straight quarters of 40%+ growth and the Street is still rubbing its eyes.
- The Legacy CDN (36% of revenue, down 7%) - Yes it's declining. No it doesn't matter as much as the ghey bears think because it's a smaller slice of the pie every single quarter while the other two run circles around it.
THE ANTHROPIC DEAL AND WHY IT'S ACTUALLY HUGE
$1.8B over 7 years. Largest contract in company history. A company that also writes checks to AWS, Google, CoreWeave, and anyone else with a spare GPU rack - looked at every cloud option on the planet and decided the CDN grandpa belongs at the table too.
Now connect that to everything above. Anthropic is getting crushed by inference costs. Memory prices have gone parabolic. Their users are hitting rate limits. They are desperately looking for infrastructure that can run their models cheaper and at greater scale.
Akamai claims you can run AI inference on their platform at up to 86% lower cost than traditional hyperscaler infrastructure. Cut that in half and call it marketing - you're still talking about running Claude at potentially 40% lower cost per token. When you're Anthropic and you're subsidizing Max plan users who are burning thousands of dollars worth of compute for $200 a month, 40% cost reduction is a lifeline.
Anthropic is doing the math on their cost structure and writing a $1.8B check to the one infrastructure provider that can actually move the needle on their inference bill.
And the implicit signal here is enormous. Anthropic doesn't hand a nine-figure-per-year contract to infrastructure that can't perform. They passed due diligence that almost nobody outside AWS, Azure, and Google has ever cleared. That alone makes it materially easier for the next ten enterprise AI buyers to pick Akamai without spending six months on evaluation. It's a sales funnel that runs itself.
One more thing from the earnings call that nobody's talking about enough - their pipeline already exceeds their current GPU inventory. They may place additional GPU orders in H2 2026 that aren't in current capex guidance. More demand than supply, with a contracted $1.8B anchor customer already locked in. They are not scrambling for revenue.
And the Anthropic deal followed a $200M deal with another unnamed US tech company signed just that February. Two deals that size in one quarter is not a coincidence.
THE NETWORK MOAT
4,300 points of presence in 700+ cities worldwide. Servers inside basically every meaningful city on earth, with peering agreements with nearly every ISP that exists - paid for over 25 years of CDN operations.
AWS can't replicate that in a decade. Cloudflare has more PoPs but optimized for lightweight stateless functions, not stateful GPU workloads. Nobody else has this exact footprint.
AI inference needs exactly this footprint. Training? Do that in a Virginia mega-campus. Running Claude for actual users in real time? You need compute near people, low latency, servers in cities AWS hasn't bothered with. That's exactly what Akamai's been building since 1998 for a reason that didn't exist yet.
Something nobody's pricing in. The semiconductor crunch actually advantages Akamai specifically. When GPU and memory costs are ripping, the operator that can run inference at lower cost per query because of network proximity, because of intelligent routing, because of a physical footprint they've already paid for - wins the next round of contracts. The labs can't just throw more expensive GPUs at the problem. They need to run what they have more efficiently. Distributed edge inference is that efficiency. Rising memory prices are Akamai's pitch deck doing the rounds on its own.
THE NUMBERS
- Q1 2026 CIS revenue: $95M, up 40% YoY
- Security revenue: $590M, up 11%
- Cash on hand: $1.73B
- Bought back $206M of stock in ONE quarter
- Gross margin: ~59%
- EBITDA margin: ~32%
- Funding the entire Anthropic capex build from operating cash flow. No equity raise. No convertibles. Actual profit money.
Compare that to neoclouds doing sale-leasebacks just to buy GPUs. Akamai is funding a war chest from operations while competitors fund theirs with vibes.
VALUATION
After the 27% rip, AKAM trades at roughly 2.7x forward sales. Cloudflare trades at ~26x. For a company with profitable operations, $1.8B of contracted backlog, and 40%+ growth in its AI segment.
Basic sum-of-the-parts:
- Security at 8x sales (cheap for a high-quality cybersecurity asset) = ~$19B
- CIS at 8-10x forward = $3-5B and compounding fast
- Legacy CDN at 1.5-2x = ~$2.5B
- Total: $24-27B vs ~$20B market cap post-rally
The stock ripped 27% and is still arguably cheap. KeyBanc raised their target from $120 to $195 in a single shot.
THE SECURITY BUSINESS IS ALSO GREAT AND NOBODY CARES
Sole Customers' Choice in Gartner's 2026 "Voice of the Customer" for API Protection. 93% customer recommendation rate. Sole. One. The only one on the list.
Why does this matter specifically right now? AI scrapers are going feral. Bot traffic is up 300% since generative AI took off. Every website is getting hammered by extraction bots, agent traffic, AI-driven zero-days. Every company needs Akamai's protection more urgently than they did a year ago, and that urgency is only growing.
The AI boom is generating customers for their security business at the same time it's generating customers for their cloud business. Double dip. Wall Street still has this classified as a media/CDN company.
THE FULL THESIS IN ONE PARAGRAPH
Semiconductor and memory costs are at historic highs and won't ease until new fabs come online in 2027 at the earliest. This makes AI inference dramatically more expensive for every lab running it at scale. Anthropic's users are getting rate limited. Anthropic is subsidizing compute and losing the pricing war with their own cost structure. Every frontier lab is looking for cheaper, distributed infrastructure that can actually move their unit economics. Akamai spent 25 years building exactly that network without knowing this moment would come. Anthropic just handed them $1.8B to prove it works. The next five labs with the same cost problem are watching. The semiconductor bull market is Akamai's sales pitch doing the rounds for free.
RISKS
- Legacy CDN decline could accelerate from -7% to -12%, keeping blended growth stuck in single digits even with CIS flying
- Capex is 40-42% of revenue. FCF is compressed for the next 18-24 months. That's real money going out the door.
- Hyperscalers can price-war. AWS, Azure, and Google don't like losing inference share and they have deep pockets.
- Anthropic at full ramp is ~6% of revenue. AI labs are volatile counterparties.
- Q2 guidance was soft. The next quarter still has to actually be executed.
BOTTOM LINE
Akamai spent 25 years building the perfect distributed computing network for a moment that didn't exist yet. Then AI inference showed up needing exactly that network. Then semiconductors went into the strongest pricing cycle in history, making inference costs a genuine crisis for every lab running at scale. Then Anthropic wrote them $1.8B to fix it.
The stock still trades like it's a dying CDN business. It's not. It's a profitable, cash-generative, globally-distributed AI inference platform that gets more valuable every single day that GPU and memory prices stay elevated - and per every semiconductor analyst alive right now, that's going to be a while.
KeyBanc says $195. Guggenheim says $181. Craig Hallum says $190. I say MOON.
Position - 47k
The poster’s core target is Akamai Technologies (AKAM). The thesis is that the market still prices it as a legacy CDN company with weak growth, but Anthropic’s 7-year, $1.8 billion cloud infrastructure contract may prove that AKAM is transitioning from CDN to an AI inference infrastructure platform, and therefore deserves a higher valuation. The poster first sets up the macro narrative: rising AI compute costs for GPUs, DRAM, HBM, and other components are putting model companies like Anthropic under pressure from inference costs and rate limits, creating a need for distributed inference capacity that is cheaper than hyperscalers and closer to users. The poster then frames AKAM’s 4,300 PoPs, coverage across 700+ cities, and long-standing interconnection relationships with ISPs as a low-latency edge inference moat. In the business breakdown, the poster emphasizes that Security accounts for about 55% of revenue and is growing 11%, Cloud Infrastructure Services accounts for about 9% and is growing 40%, while legacy CDN accounts for about 36% and is declining 7%, arguing that the shrinking business is becoming less important while security and cloud are reshaping the company. On valuation, the poster says that even after AKAM surged 27% in one day, it still trades at about 2.7x forward sales, while Cloudflare (NET) trades at about 26x, arguing that the gap is too large. Using a rough SOTP, the poster assigns Security 8x sales, CIS 8-10x sales, and legacy CDN 1.5-2x sales, arriving at a value of $24-27 billion versus a market cap of about $20 billion, implying further upside. The image shows the poster holding 320 shares of AKAM at an average cost of $140, with the stock at $148.50 at the time of the screenshot, market value around $47,520, daily gain around $10,179 or +27.26%, and total gain around $2,720 or +6.07%. This is consistent with the text saying “Position - 47k,” indicating this is not pure talk, but the position is also highly exposed to short-term volatility. Catalysts include the Anthropic contract beginning to ramp, follow-on purchases from other AI labs or enterprise buyers, continued high growth in CIS, analyst target price upgrades, and AI bots/scrapers driving demand for the security business. The most important implicit assumptions are: Anthropic chose AKAM not as a temporary compute supplement, but as validation of AKAM’s cost/performance advantage in large-scale inference; edge PoPs will become important enough for LLM inference; AKAM can absorb high capex through operating cash flow; and CIS’s high growth can offset CDN decline. The main weaknesses are: the poster extrapolates Akamai’s official “up to 86% savings” from a specific case to Claude-scale inference without sufficient evidence; while multiple media outlets have reported the Anthropic contract customer to be Anthropic, the company’s announcement described it as a “leading AI/frontier model provider,” and contract concentration and execution risk are high; whether AI inference truly needs city-level edge nodes was also challenged in the comments; additionally, AKAM has already risen sharply within a week, so the near-term valuation reset may already be partially reflected.
The overall comment sample is not large, and the post score is only 18 with an upvote ratio of 0.63, indicating that the community has not formed a strong consensus. Supporters mainly agree with the asymmetric re-rating logic of “AKAM valuation far below Cloudflare + $1.8 billion Anthropic contract + 40% CIS growth.” Some said they took another serious look after seeing the valuation gap, while others argued that if AKAM becomes a discounted inference backend under GPU/HBM cost pressure, then it is no longer just an old CDN. Another commenter claiming to be a long-term Akamai user added that Akamai is very good at embedding large customers into its full technology stack and signing multi-year large contracts, with high enterprise migration costs; even if the Anthropic relationship changes, AKAM’s existing business would not disappear. Opposition and skepticism focus on three points. First, a highly upvoted comment says the claim that “nobody is talking about it” is not valid, because AKAM has already risen sharply year-to-date and within the past week, making momentum-chasing risk obvious. Second, some users questioned the actual value of PoPs for AI inference, arguing that ordinary LLM users do not care much about a 5ms versus 200ms difference, and that large cloud providers’ continent-level deployments are already sufficient; the advantage of edge city-level nodes may only be more apparent in voice AI, multi-turn agent interaction, or local/low-latency scenarios. Third, some worry that Anthropic itself is burning cash, rate-limited, and has unsustainable pricing strategies, even arguing that if Anthropic cancels the partnership, AKAM’s stock price would pull back violently. The poster responded relatively little, mainly replying to someone who said they would buy puts by saying the chart was a multi-year breakout and they did not want to short it. When challenged on the claim that nobody was talking about it, the OP only said they had not seen much discussion on WSB, which did not resolve the “already priced in by the market” issue. Another highly upvoted comment suspected that the post and comments looked like they were written by AI or bots, warning that sentiment in this post may be amplified by automated content. The real meaning of these memes and sarcasm is that the community remains cautious about “AI narrative packaging around an old company.” Overall, the comments acknowledge the cheap valuation and the importance of the contract, but show clear disagreement over the edge inference moat, the sustainability of the Anthropic contract, and the risk of buying after a major rally.
Loading Truth Social...
Loading live updates...