Fastly is a Nasdaq-listed edge cloud infrastructure company whose revenue has accelerated and whose non-GAAP operating income turned solidly positive, yet the stock still prices as a loss-making startup. The second quarter 2026 results mark the inflection: GAAP gross margin reached a record level, and non-GAAP operating income swung to a positive figure from a loss a year earlier, while the company raised full-year revenue guidance to the midpoint of roughly $739 million.
The mechanism driving the shift is a structural repricing of Fastly's cost base as its Compute platform matures and its security business, which grew 43% year over year, scales on lower marginal infrastructure. The last twelve month net retention rate climbed to 117%, the highest level in over three years, which means existing customers are spending materially more each year and the revenue base compounds rather than resets. That combination of expanding gross margin and accelerating retention is what the prior pricing structure did not account for.
The tension is customer concentration and the timing of the convertible notes. One customer accounted for 11% of second quarter revenue, and the top ten customers together represented more than a third of the total. That concentration signals deep enterprise penetration but also makes the revenue stream vulnerable to a single procurement decision. The convertible notes, with roughly $362 million of principal outstanding, sit as a refinancing event that management has already begun working through with partial repayments in the first half of 2026.
The catalyst is the third quarter report, due in early November, which should confirm whether the raised guidance is a floor or a starting point. With full year non-GAAP operating income guided at a midpoint of roughly $92 million, even a modest beat would put Fastly on a credible path to sustained profitability by the second half of 2027.
Fastly operates a programmable edge cloud that sits between end users and origin servers, processing web traffic, API calls, and increasingly AI generated workloads at roughly 100 points of presence across the globe. The company describes its platform as the convergence of a content delivery network with the functionality of hardware appliances that enterprises previously bought separately: application delivery controllers, web application firewalls, API management, bot detection, and distributed denial of service protection. That convergence is the core of the strategic argument, because it means a customer that once paid three or four vendors for overlapping edge functions can consolidate them onto a single software platform with a single bill and a single set of deployment APIs.
The competitive frame is defined by the two giants. Cloudflare and Akamai dominate the CDN layer by scale, and both have spent a decade building their own security and compute offerings. Fastly's wedge has always been developer experience and programmability: the Compute platform lets engineers deploy serverless code in the same seconds that a CDN refreshes a cache, and the company's Rust based runtime gives it a performance profile that hardware appliance vendors cannot match. In 2026, that wedge has widened. The company published research showing that AI generated traffic is growing roughly 6.5 times faster than human traffic, and it has positioned the Compute platform as the natural home for that machine traffic. The partnership with Skyfire, announced in the second quarter, extends the platform into agent to agent commerce, letting enterprises verify and transact with AI agents in real time at the edge. That is a concrete step into a market that did not exist three years ago.
The go-to-market motion has shifted from a bottom-up developer adoption model to a top-down enterprise sales engine. The company has built out a field sales organization that targets the top ten customers now representing 37% of revenue, and the last twelve month net retention rate of 117% reflects expansion within that base rather than logo count growth. That model is more predictable than the developer-led motion it replaced, but it also makes the revenue stream more dependent on a smaller number of large procurement decisions. The company's fiscal 2025 annual report discloses a substantial goodwill balance from prior acquisitions, a reminder that Fastly has grown partly through bolt-on deals in the security and media space, and those acquired businesses contribute a meaningful share of the security revenue line that grew 43% in the second quarter.
The Nasdaq listing transfer, completed in December 2025, is a small but telling signal. Fastly moved its shares from the NYSE to the Nasdaq Global Select Market, a venue with a higher concentration of growth and technology investors. The company cited access to its target investor base as the reason, and the transfer coincided with a period when the stock had been trading at a significant discount to its growth trajectory. Whether that repositioning is a catalyst or a formality remains to be seen, but the optics matter for a company trying to re-rate its multiple.
The platform rests on three product families. Network Services, which generated roughly $134 million in the second quarter, covers the core CDN and delivery function: caching, load balancing, image optimization, and the programmable compute runtime that lets customers deploy code at the edge. Security, at roughly $42 million, bundles the web application firewall, bot mitigation, DDoS protection, and API security into a single subscription. The third line, Other revenue, carries the Observability product and the residual compute offerings that have not yet been reclassified. The revenue mix is shifting toward the higher margin security and compute lines, and that mix shift is the primary driver of the gross margin expansion.
The Compute platform is the moat that the competitors have not yet closed. Fastly's runtime is written in Rust and executes customer code in isolated, sub-millisecond sandboxes at the edge, a performance profile that the JavaScript based runtimes used by Cloudflare Workers cannot match for latency sensitive workloads. The company released a C++ SDK in the second quarter of 2026, extending the developer surface to low level AI inference and gaming workloads that demand the tightest possible latency budgets. That is a deliberate play to own the edge layer of the AI inference stack, where the company argues that routing model calls through the closest point of presence reduces round trip time by enough to matter to end user experience.
The second quarter partnership with Skyfire is the most concrete expression of that AI positioning. Skyfire provides an agent identity and verification layer that lets enterprises confirm the authenticity of AI agents before they transact, and Fastly's role is to run that verification at the edge so the check happens in milliseconds rather than seconds. The strategic logic is that as AI agents begin to buy, sell, and negotiate with each other, the edge becomes the natural enforcement point for trust, and the company that controls the enforcement point captures the transaction fee. That is an early stage market, and the revenue contribution in 2026 is negligible, but the partnership signals where the company believes the next revenue curve originates. The LALIGA collaboration, announced in the same quarter, addresses a different pain: illegal streaming of live sports. The anti piracy solution detects and blocks unauthorized redistribution of premium sports content at the edge, and it is aimed at the rights holders who have been losing significant revenue to pirate streams. The partnership matters less for immediate revenue than for the reference value it provides: live sports is the most demanding use case for real time edge delivery, and a marquee logo in that space validates the platform's ability to handle the highest concurrency and lowest latency workloads in the AI inference pipeline.
The network itself is a modest moat. With 578 terabits of connected capacity and roughly 100 points of presence, Fastly's infrastructure is large enough to serve enterprise workloads globally but an order of magnitude smaller than Cloudflare's footprint. The company offsets that scale gap with the performance advantage of the Compute runtime and with the consolidation argument: a customer that replaces a CDN, a WAF, and an API gateway with a single Fastly contract saves on license fees, integration labor, and operational overhead. That is a stickier economic relationship than a pure CDN subscription, and it is why the net retention rate is expanding rather than flat.
Full year revenue reached $624 million in 2025. That represented a 15% increase over the prior year. The second quarter accelerated that pace to 23% growth at $183 million. Network Services, the core delivery and compute line, grew 17% in the second quarter. Security grew 43% and the Other line, which carries Observability, grew 69%. The faster growing lines are also the higher margin lines, and that mix shift is the mechanical reason the gross margin expanded.
Gross margin is the number that changes the thesis. GAAP gross margin reached 63.3% in the second quarter, a record for the company, compared with 54.5% a year earlier. The non-GAAP figure, which excludes stock based compensation and amortization, was 65.8%, up from 59.0% in the same quarter of the prior year. The operating loss narrowed to $14.4 million on a GAAP basis, down from $36.9 million in the prior year quarter. Non-GAAP operating income swung to a $27 million profit from a $4.6 million loss.
The cash flow profile improved at the same time. The operating cash flow for the first half was $68.2 million. That was nearly 60% higher than the $43.1 million in the same period of the prior year. Free cash flow turned positive at $3.6 million in the second quarter after being positive in the first as well. The balance sheet carries a meaningful liability. The company had roughly $362 million in cash and current marketable securities as of year end.
The mid-2026 position had shifted to $89.8 million in cash. Marketable securities stood at $247.7 million. That represented a roughly $337 million liquidity buffer. Against that, the long-term debt balance stood at $323.3 million at year end. The company made partial repayments of the convertible notes during the first half, reducing the current portion to $38.6 million. The remaining principal on the notes is the refinancing event that hangs over the second half of the year. The income statement below the operating line is cleaner than it has been in years. The GAAP net loss for the full year narrowed to $121.7 million from $158.1 million a year earlier. The company has been modestly dilutive through the exercise of options and the employee stock purchase plan, but the rate of dilution is slow relative to the revenue growth. The stock based compensation expense, which is the primary bridge between GAAP and non-GAAP results, is stable rather than growing with the revenue base.
Management raised full year revenue guidance to a range of $732 million to $746 million. Non-GAAP operating income guidance rose to a midpoint of roughly $92 million, a meaningful lift from the prior range. The midpoint implies roughly 18% growth for the full year, and the operating income number would put the company on a run rate that supports a credible path to GAAP profitability by the second half of 2027. The guidance is the first number that matters, and the third quarter report in November is the test: if revenue lands at the top of the range, the full year number is essentially locked and the stock should re-rate.
The execution risk is concentrated in two areas. The first is the customer concentration that has been rising in parallel with revenue growth. The top ten customers represented more than a third of second quarter revenue, and a single customer accounted for 11%. That is a sign of deep enterprise penetration, but it also means that a procurement decision at one large customer can move the quarterly revenue line by more than a point of growth. The company's go-to-market engine is built to offset that risk through net expansion at existing accounts, and the 117% net retention rate is the evidence that the offset is working, but the concentration trend is something to watch in the next two reports.
The second execution risk is the refinancing of the convertible notes. The notes carry a 0% coupon, so there is no interest expense to service. The principal is settled in cash or shares at maturity, and the company has been making partial repayments in the first half, reducing the current portion to $38.6 million. The remaining long-term balance of roughly $323 million still needs to be addressed. With a liquidity buffer of approximately $337 million in cash and marketable securities, the company has the cash to retire the notes without a dilutive equity raise, but the decision of whether to convert, refinance, or repay determines the equity value that flows to existing shareholders in the second half of the year.
The broader execution question is whether the AI traffic thesis converts into revenue. The company has published research showing that AI generated traffic is growing 6.5 times faster than human traffic, and it has positioned the Compute platform and the Skyfire partnership to capture that growth. The revenue contribution from AI workloads in 2026 is not separately disclosed and is embedded in the Network Services and Other lines. The risk is that the AI traffic wave reaches the CDN layer before it reaches the compute and security layers where Fastly has the differentiation, and the company's competitive position depends on whether the inference and agent workloads arrive at the edge on Fastly's infrastructure rather than on the hyperscaler networks where the model providers are already hosted.
The largest single risk is a loss of one of the top ten customers. With those ten accounts representing 37% of revenue and the largest at 11%, a procurement decision at a single enterprise can remove roughly a point of quarterly revenue growth overnight. The mechanism is straightforward: the customer's CTO or head of infrastructure decides to migrate workloads to a hyperscaler's native edge service or to Cloudflare's bundled offering, and the revenue line drops by the amount of that customer's spend. The company's mitigation is the net expansion at existing accounts, which is working at 117% retention, but a large account loss would require two or three quarters of new logo growth to replace, and the stock would likely de-rate on the announcement.
The competitive risk is the hyperscaler encroachment. Amazon, Google, and Microsoft all operate edge delivery networks that are adequate for standard CDN workloads, and their bundled pricing against customers who are already paying for compute and storage on those platforms makes the marginal cost of adding delivery a non event. Fastly's defense is the performance advantage of the Rust based Compute runtime and the consolidation of multiple edge functions into a single contract, but that defense erodes if the hyperscalers close the performance gap on the compute layer. The company's published research on AI traffic growth is an attempt to reframe the competitive set, arguing that the next wave of edge workloads is not standard CDN but AI inference and agent traffic, where the hyperscalers' own model hosting creates a conflict of interest that keeps inference workloads on third party infrastructure. That argument is plausible but unproven, and the revenue data in the next two quarters is the test.
The balance sheet risk is smaller than the operating risks but worth naming. The convertible notes carry a 0% coupon, so there is no interest expense to service. The principal balance of roughly $323 million in long-term debt represents a refinancing event that the company can fund from its $337 million liquidity buffer without a dilutive raise. The downside scenario is a conversion at a stock price below the conversion threshold, which forces a cash repayment and reduces the liquidity buffer below a comfortable level. The company's partial repayments in the first half show that management is working to reduce the principal ahead of the maturity date, and the absence of a coupon means there is no cash drain from the notes themselves. The risk is the timing: if the stock stays depressed into the second half of the year and the company elects to repay in cash rather than refinance, the liquidity buffer shrinks at the same time that capex spending on network expansion is accelerating.
The counterargument to the bull case is that the gross margin expansion is a one time event driven by the mix shift toward the already profitable security line, and that the underlying CDN economics remain structurally unprofitable at scale. The evidence against that is the direction of the net retention rate, which is expanding rather than flat, and the fact that the operating cash flow is growing faster than revenue. But the counterargument has merit in the sense that if the security growth slows and the Network Services line reverts to its historical margin profile, the non-GAAP operating income number that the company guided to $88 million to $96 million for the full year could compress meaningfully, and the stock would de-rate to a multiple that reflects the underlying CDN business rather than the security and compute story.
At the current price of roughly $20.60 per share, Fastly carries a market capitalization of about $3.2 billion. The share count stands at approximately 157.6 million. The trailing twelve month revenue is $624 million on a full year basis, though the first half has already put the company on a run rate that exceeds that. The enterprise value, after adjusting for the long-term debt and the liquidity buffer, lands at roughly $3.2 billion, which puts the enterprise value to trailing revenue multiple at about 5.1x. That is a discount to the 6x to 8x range that Cloudflare trades at on an enterprise value basis, and the discount reflects the smaller scale, the higher customer concentration, and the history of GAAP losses.
The forward multiple is where the valuation argument turns. On the full year revenue guidance midpoint of $739 million, the enterprise value to forward revenue multiple compresses to roughly 4.3x. The non-GAAP operating income guidance implies an enterprise value to EBITDA multiple of approximately 33x to 36x, a number that is expensive in absolute terms but consistent with a company that is transitioning from a loss making growth profile to a profitable one. The bear case prices the stock at a multiple that reflects the CDN business only, excluding the security and compute lines, and that scenario produces an enterprise value to revenue multiple of roughly 3x, which is where the downside sits. The base case assumes the guidance is met and the non-GAAP operating margin holds at roughly 12% to 13% of revenue. That supports a multiple of 4.5x to 5x forward revenue. The bull case assumes the AI workloads begin to contribute meaningful revenue by the second half of 2027 and the net retention rate continues to expand, which supports a re-rate to 6x forward revenue and a multiple that reflects the compute and security mix rather than the legacy CDN profile.
The most useful valuation frame is the net retention rate as a leading indicator. A company with 117% net retention and expanding gross margin has a revenue base that compounds rather than resets, and the historical multiple for companies at that inflection point is higher than the current price reflects. The 0% coupon on the convertible notes is a structural advantage that keeps the enterprise value clean, and the absence of a large interest expense means that the path from non-GAAP profitability to GAAP profitability is a function of stock based compensation and depreciation rather than financing costs. The stock based compensation, at roughly $85 million on an annualized run rate, is the primary gap between the two numbers, and it is stable rather than growing with the revenue base.
The valuation conclusion is that the current price embeds a bear case that assumes the margin expansion is temporary and the revenue growth decelerates. The evidence in the second quarter report, the raised guidance, the expanding retention rate, and the positive free cash flow, argues that the bear case is the minority scenario. The stock is not cheap on an absolute multiple basis, but it is cheap relative to the trajectory of the underlying business, and the re-rate opportunity is concentrated in the two quarters before the convertible notes maturity, when the market is forced to price the refinancing outcome and the path to sustained GAAP profitability.
Fastly sits at an inflection point that the market has not fully priced. The second quarter results, the raised full year guidance, the record 63.3% gross margin, and the 117% net retention rate together describe a company that has solved its margin problem and is now compounding revenue at an accelerating rate. The stock at roughly $20.60 prices a business that is still loss making on a GAAP basis and that carries a customer concentration risk that is real and rising. That is the honest bear case, and it is the case the market is buying.
The bull case is that the margin expansion is structural rather than one time, driven by the mix shift toward the higher margin security and compute lines and by the operating leverage of a platform that has already built out its global infrastructure. The non-GAAP operating income of $27 million in a single quarter, the positive free cash flow, and the $337 million liquidity buffer that covers the convertible notes without a dilutive raise, all point to a company that is close enough to sustained GAAP profitability that the next two quarters define the outcome. The AI workloads thesis, the Skyfire partnership, and the C++ SDK release are early stage options on a revenue curve that does not yet appear in the income statement, and they are the reason the stock has a re-rate path that the bear case does not capture.
The judgment is that the current price is below the value of the business on a forward basis, with the asymmetry tilted toward the upside over a twelve month horizon. The catalysts are sequenced: the third quarter report in November confirms the guidance trajectory, the refinancing of the convertible notes in the second half of the year removes the overhang, and the AI revenue contribution, if it materializes by the first half of 2027, justifies a higher multiple. The risks are real, the customer concentration is the most immediate, and the hyperscaler competition is the most durable. But the direction of the underlying metrics, the retention rate, the gross margin, the operating cash flow, all point in the same direction, and that alignment is what makes this a higher conviction position than the multiple alone would suggest.