Archives

Google has been pretty quiet about its Media CDN offering, but that’s changing as the company has signed multiple large M&E customers and added more capacity and new functionality. The company has allowed me to disclose that Google Media CDN has over 100Tbps of egress capacity and growing. Not to be confused with Google Cloud CDN, which is for web acceleration, Google’s Media CDN service leverages the same infrastructure that YouTube uses to serve its global users.

Google’s Media CDN service targets customers with high-volume media and large file downloads, as well as support for live streaming when it makes sense. Google’s support for live video prioritizes customers who commit to a specific level of sustained traffic. This is a smart business decision since even the largest live events when measured by concurrent users, don’t drive much revenue but consume a lot of network capacity. Google targets customers prioritizing low latency for live, high origin offload for VOD, and maximum throughput for downloads. For VOD, Google says their cache hit ratios average ~98.5-99% and for live, ~96.5-98.5%.

Google Media CDN leverages Google’s edge network, which is present in over 3,179 locations worldwide and offers origin protection by default. This doesn’t mean that Google’s Media CDN service is deployed in 3,000 locations, but the Media CDN team can add capacity across specific locations based on customer needs, leveraging Google’s massive network footprint. Google Media CDN also supports edge programmability, enabling custom logic for services like header manipulation, token authentication etc., directly within the request path.

Later this quarter, Google Media CDN will add Flexible Shielding, enhancing flexibility for customers with centralized origins serving global audiences. This service allows customers to configure specific shield node locations for optimized performance and offload. Customers can also add Google’s Cloud Armor service for WAF and DDoS protection at the edge. Google Media CDN offers near real-time logging and metrics integrated into Google’s Cloud operations suite (Cloud Logging, Cloud Monitoring) and provides ready-to-use dashboards. Its metrics include server-side metrics (TTFB, TTLB, Cache Status, Origin Latency) and client-side insights via full support for CMCD.

Google shared a list of client names using their Media CDN for streaming and software downloads, but I don’t have permission to mention most of them. I previously disclosed that WBD uses Google Media CDN in their multi-CDN deployment, and last week, Google added a testimony with MLB on their use of Google Media CDN, which you can see here. Google already has relationships with large M&E customers using Google’s Cloud stack, so over time, I expect its Media CDN offering to be added into the mix for customers using a multi-CDN approach for streaming and large object downloads.

Varnish Software has launched a new private CDN offering called Ora Streaming with its long-time partner, Intel. Fully managed by Varnish, with direct access to the resources that develop Varnish Enterprise, the proprietary version of Varnish Cache that has enhanced performance, security and functionality. Ora Streaming Inc. (incorporated in Delaware) is 100% owned by Varnish Software AB (registered in Sweden), the parent company in the Varnish Software group. Initial use cases for Ora Streaming are live and VOD video streaming, large file delivery, and web and API traffic, although the primary commercial focus will be video.

The new solution is based on dedicated bare metal servers and contracted capacity, not shared, for content delivery on infrastructure owned by major carriers/ISPs. Ora Streaming and Intel have formed a strategic partnership with several global tier-1 ISPs and regional tier-1 carriers targeting regional content providers. Ora shared the names of nine carriers with me, but they only have permission to share the names of Lumen and Telekom Malaysia at launch. The tier-1 ISPs have a leveraged position to help establish a dedicated peering capacity with last mile carriers. Ora Streaming’s strategy is to continue developing carrier-neutral footprints augmented with regional tier-1 ISPs.

Ora Streaming’s current global capacity is an aggregate combination of edge bare-metal and egress capacity available for deployment across all its global partners. The approximate global peak capacity is 30-35 Tbps, which can increase to 50 Tbps based on availability. Those numbers are aggregate capacity available through Ora’s partners at a given time but not deployed Ora capacity.

Varnish says its initial customer target is streaming providers in North America, Europe, East Asia, Japan, Australia and South Africa. The company says Ora Streaming becomes economical for customers at a minimum delivery of 20 PB/month and that they have identified 250 target accounts in its target markets that fit this profile. Customers pay a fixed monthly fee based on capacity with no additional cost for egress, traffic or overage fees. In addition, edge compute workloads are included in the fixed monthly fee so customers can execute programmable workloads like ad insertion, log management, and traffic shaping directly at the edge. The company’s website says Ora Streaming has a lower delivery cost, but they didn’t discuss pricing with me, and it’s unknown what they are comparing it to.

As Ora is based on Varnish Enterprise, an extensive library of Varnish VMODS can tailor the private CDN to specific customer requirements. In addition, the Varnish Configuration Language (VCL) can be used to its full extent, without limitation, which differs from other Varnish-based CDNs, where VCL is restricted to avoid noisy neighbors or even outages on the shared platform. Ora Streaming launches with a lot of functionality tied to access control, connection encryption, content protection and connection authentication. Ora Streaming is currently in preview mode by invite only.

Varnish and Intel will speak at the NAB Streaming Summit on privately managed CDNs, discussing enabling real-time configuration and supporting advanced use cases such as ad insertion, ad stitching, watermarking, and custom log generation at the edge. You can see more details on that session here.

For the final NCAA March Madness game available on Max Saturday March 29th, I did a quick test of the stream across nine platforms and looked at the quality of the stream, encoding bitrates, key moments feature and CDNs used. It was an excellent experience with fast startup times, no rebuffering and a great picture quality across all the devices I tested.

For Max’s encoding bitrate ladder, the SDR stream maxed out at 10Mbps and the HDR stream at 8Mbps. They used HLS and DASH, sharing the same underlying CMAF media segments. Chunked encoding came in at 3.2 seconds, and the stream was delivered via Amazon CloudFront, Akamai, Fastly and Google Media CDN. Warner Bros. Discovery doesn’t break out streaming viewership after live events, so it’s unknown what the AMA or concurrent stream count was. I would not expect the stream to have broken into the top 20 largest streaming events of all time, meaning less than 10 million concurrent viewers. Live streams of March Madness were available to all Max subscribers in the US, with Max having pulled live streaming from its ad-supported tier on March 30th. B/R Sports and CNN content are available at no additional cost to subscribers in its standard and premium tiers.

Individual games offered Dolby Vision, Atmos and a key moments experience functionality where you could jump to the best plays on both live and replay. Max was the only service that offered March Madness with Dolby Vision. Video startup times were all under 2 seconds, with some under 1 second, depending on the platform, with Roku always being the slowest for every streaming service I’ve ever tested. I did not do any streaming latency testing when compared to OTA, as CBS is the only network offering OTA, and Max doesn’t carry CBS games.

Max offered a multiview experience for March Madness, which they first introduced doing its NASCAR Driver CAM coverage. It was without the leaderboard and enhanced with the ability to enter and exit the full-screen view of individual views from within the multiview. Users could double-click on any of the displayed feeds, allowing the selected game to expand into full-screen mode, effectively switching to that game’s direct feed. The multiview option was accessible across all platforms, while the capability to expand a multiview feed to full screen being available on desktops and all CTV devices. I didn’t get a chance to test multiview since there was only one game left on Max, but I did previously write a little about Max’s NASCAR multiview functionality last month.

In calendar Q4 2024, Max gained 6.4 million subscribers, bringing its total to nearly 117 million. CEO David Zaslav said Max is on track to have “at least 150 million subscribers by the end of 2026.” For full year 2024, WBD’s DTC business turned a profit of $677 million, compared to a 2023 profit of $103 million.

I tested the stream via Max’s apps on LG, Samsung, Vizio and TCL/Roku TVs and streaming devices, including Apple TV, Roku Stick, Fire TV Stick, MacBook (browser), iPad and iPhone. I tested via two ISPs, Verizon and Optimum, at half-gig speed. All devices were connected via ethernet, except iPads and iPhones, which were connected via Wi-Fi. Devices used included Roku Streaming Stick 4k (3820), Apple TV 4K (A2843), Amazon Fire TV Stick 4K (M3N6RA), Amazon Fire TV Stick 4K Max (K3R6AT ), LG OLEDs (55C9AUA), Samsung TV (QN65S90CAFXZA), Vizio TV (V4K55M-0801) and Roku TCL TV (50R4A4).

At the end of 2023, privately held Bitmovin allowed me to disclose that revenue was in the “Twenties Million” range for the year. Bitmovin says that at the end of 2024, the company reached free cash flow and grew revenue by 20%. This puts Bitmovin 2025 revenue in the $30M+ range and growing. Bitmovin’s collaboration with cloud partners has been a significant accelerator, with the company being deployed across the AWS Marketplace, Google Cloud Marketplace, and Microsoft Azure Marketplace. The company has also cut many deals with end-to-end and OVP partners who white-label/OEM their products and pitch new customers together, allowing them to win business that Bitmovin alone would not be a fit for.

From the Azure Media Services migration, Bitmovin says it received 400 new logos and has expanded its reach with Tier-1 customers in the US and globally. While its business grows, Bitmovin said cost pressure increases and average deal sizes are still decreasing, with the media industry still under a lot of pressure to do more with less. Bitmovin grew its VOD encoding volume by 40%+ in 2024 and is its largest product by revenue. In 2025, it is doubling down on VOD encoding and investing in AI session interpreters, AI contextual advertising, AI-based encoding job generation and an AI sign-language avatar for its players.

Bitmovin says its focus in 2025 is similar to last year, emphasizing being more efficient with its go-to-market sales and marketing functions, improving margins, investing more in R&D, and improving both EBITDA and cash flow, which are positive. Bitmovin’s last round of funding was $25 million in 2021, and they have raised $68 million to date.

Technical challenges are a current reality for live streaming events, which is magnified even further by events with a large concurrent stream count. Load testing is complex, and troubleshooting in real-time is challenging, even for the best engineering teams. I recently talked about how, on average, 5-10% of streaming subscribers are experiencing poor QoE (a conservative estimate). This number equates to unvalidated content performance metrics, or “blind spots,” most common in harder-to-reach ISP networks. These networks are typically multiple hops away from major CDNs, meaning streamed content takes a much more complex path to reach viewers while providing much less (e.g., little to no) QoE visibility. If any geographical area or subscriber demographic is more likely to experience live-streaming issues, it’s in this blind spot.

On February 6th, the U.S. Bankruptcy Court approved the sale of select Edgio assets to Parler for $7.5 million. Parler is keeping the EdgeCast brand, and Parler Cloud Technologies is being renamed to EdgeCast Cloud Services. The company has been working to restart a portion of the original EdgeCast network, which will go live on April 15th with 10Tbps of network capacity from within 25 PoP locations, primarily for customers who need video-on-demand delivery and software downloads. At launch, live streaming will have limited functionality.

The company plans to have 50Tbps of network capacity by the end of this year, close to the approximately 65Tbps capacity the network had when it was shut down under Edgio. Under the previous owners, EdgeCast’s network hit a peak capacity of 140Tbps. EdgeCast Cloud Services will offer delivery capabilities internationally, excluding the Middle East, South Asia, Eastern Europe, Africa and Australia. Part of the capacity on EdegCast Cloud will be used for Parler’s services, consisting of its social media platform Parler and video streaming platform Playtv.

During Edgio’s Bankruptcy auction, Parler acquired EdgeCast-only gear in Ashburn, Chicago, Dallas, Los Angeles, New York, San Jose, Seattle, Miami, Washington, DC, Bogota, Buenos Aires, Sao Paulo, Amsterdam, Frankfurt, Tokyo, and Singapore. In addition to hardware assets, Parler also acquired the EdgeCast software stack, including WAF and other non-CDN services, EdgeCast ASNs and 6 IP blocks. It did not acquire anything related to Edgio’s Limelight brand or assets. Since acquiring the assets, EdgeCast Cloud Services has hired approximately 100 former Edgio employees, 70% of whom are related to engineering and operations. The employees hired were not part of the bankruptcy process.

EdgeCast Cloud Services is powered by Triton, formerly Joynt Legacy Cloud Software, which originated as Samsung Cloud. Parler says that when combined with its open compute hardware, the integration of the Edgecast CDN enhances its cloud offering, transforming it into a full-stack cloud services platform. The parent company, Pulse, is building a decentralized digital ecosystem spanning social media, streaming, payments, e-commerce, cloud services, blockchain, and content delivery.

While the company discussed its 2024 revenue and growth numbers with me, I don’t have permission to release them. The company said over the next few months, news will come out from Pulse regarding funding and additional acquisitions.

On Akamai’s Q4 2024 earnings call, the company provided an update on its delivery business for 2025 and beyond. Edgio contracts contributed $9M of revenue in Q4 2024 and full year 2025, Akamai expects $85M-105M in revenue from Edgio customer contracts. Akamai said its capex spend will increase by approximately 1% of revenue to accommodate the increased traffic resulting from the Edgio transaction. “Over 100” of the Edgio contracts were new customers for the company, and Akamai does not expect “significant churn” of the customers.

Akamai noted they are “beginning to see signs of improvement in the delivery marketplace. With more customers willing to sign multiyear contracts and with predictable pricing. A more stable pricing environment generally and early signs of stabilizing traffic growth.” If these trends continue, Akamai expects its delivery revenue to decline 10% YoY in 2025 versus the 15% decline it saw in 2024. And if the positive trends continue, Akamai expects the decline to shrink even further in 2026 and beyond. For the full year 2024, Akamai’s Delivery revenue (not CDN) was $1.31B, down 15% year-over-year. Akamai’s Delivery revenue by year:

• 2024: $1.31B
• 2023: $1.54B
• 2022: $1.66B
• 2021: $1.87B

Akamai mentioned that their largest customer is navigating “political challenges” in the US and pursuing a DIY (CDN) strategy. While Akamai didn’t mention the customer by name, it’s TikTok. Both of these factors could impact Akamai’s revenue. The good news is that in Q4, Akamai signed a 5-year contract that includes a “substantial minimum annual spend,” which provides greater predictability and reduces Akamai’s exposure to the customer’s political situation in the U.S.

Akamai paid $125M for select customer contracts from Edgio’s bankruptcy. In addition, they spent $25-$30M on transition service costs, so Akamai spent $150-$160 million on the deal. Based on their guidance of $85M-105M in revenue from Edgio customer contracts in 2025, it appears the contracts Akamai acquired will pay off the Edgio deal by year two.