Welcome to this edition of our Weekly Analyst Insights roundup, which features key insights that our analysts have developed based on the past week’s events.
The other day we published a Forbes article I’m really proud of, encapsulating what Qualcomm is attempting to do with its disciplined approach to diversification across end markets. The piece is based on exclusive one-on-one interviews I did with four of the company’s top operating executives, all of whom have been with Qualcomm for 20-plus years. While working to maintain its leadership position in mobile handsets, Qualcomm is continuing a decade-plus of effort in automotive chips, bolstered by its expanding work in IoT, PCs, the datacenter, and — on the horizon — robotics.
For now, the main growth businesses may continue to be automotive and IoT, which have been performing strongly for a while. (See my writeup from the end of last year about that.) But in the long run, the bigger opportunities might be in the datacenter and robotics, where Qualcomm’s specialty in delivering high compute performance with the lowest power consumption will continue to be a competitive differentiator.
Check out this week’s Analyst Insights roundup for more from the MI&S team, including what’s top of mind for each analyst, our thoughts on vendor announcements, press quotes, and more.
Last week, President Trump’s call for Intel CEO Lip-Bu Tan to resign was top news; this broadcast analysis of my appearance on CNBC summarizes the issues at hand. MI&S analysts were also quoted in top-tier global publications, including the BBC, Fierce Networks, Business Today, Investors Observer, Mobile World, The Times, and Reuters. I was also a guest on The Prof G Pod to discuss AMD’s recent earnings.
This month, our analysts are attending the HP Summer Gaming Event, Hot Chips, VMware Explore, the US Open & IBM Innovation Studio, and the T-Mobile Technology Innovation Summit 2025. We’re all preparing for a very busy fall event season. See the full calendar of events we’re attending and reach out if you’ll be there too. We always love to see your smiling faces IRL! Until then, follow Moor Insights & Strategy on X, LinkedIn, and Instagram for live updates and hot takes throughout the week.
Our MI&S team published 17 deliverables:
See more in the full Analyst Insights.
Have a great week!
Patrick Moorhead
MI&S Quick Insights
OpenAI had a monster week. And kudos to its marketing team for taking advantage of the slow summer months to get all eyes on the company when competitors largely stay dormant. The biggest news was GPT-5; after a two-year wait, we now see a new model that is a big step forward from its predecessor. While it did post impressive bench scores, in some ways GPT 5 is catching up with and not outpacing Claude, Gemini, and even in a couple of cases Grok. But while we may not see earth-shattering performance, we do see a new level of richness and breadth from the model, and I expect it will once again be the general-purpose AI standard for some time.
The day before GPT-5 was announced, OpenAI also pushed out two new models as open source. This was a big step for the company, as open APIs and not models have been its mode of operation. But what may be more notable than the strategic pivot is that these models — despite being hyped as available on all the major clouds on day 1 — are targeting PCs and mobile devices. This will surely open up some new doors into edge use cases and possible hybrid AI scenarios. For example, a local model could be a constant context collector based upon user inputs and actions, which could then feed into larger cloud-based models for heavier reasoning exercises. Imagine a situation where your phone is able to adjust the settings for any car you get into, even if you have never sat in it before. So, while the clouds will sell instances with these smaller models, I am way more interested to see what happens outside the cloud with these new models (as well as with Llama and others that compete with it).
Despite all of this progress with models, I do think that we are starting to see a plateau in terms of AI capabilities for enterprises. The missing piece will be the more effective use of corporate data to drive more effective context into models that are already pretty good. Lately, I have been turning my attention to DevOps and data pipelines, where I think the next AI breakthroughs will happen.
Last week, I had the chance to get briefed by Snowflake about its Cortex AI capabilities, and I learned quite a bit. While the ability for AI to generate SQL data queries is a key feature for business analysts, I was just as impressed with how Cortex AI was able to present outputs in text, tabular, and graphical formats. It also helps that Snowflake is data-agnostic and cross-cloud.
Cortex AI reinforces my belief that we are approaching a new phase in the evolution of enterprise AI. Even though experts have been saying that great data is the foundation for great AI since the advent of large language models years ago, we are only recently seeing that notion take hold in enterprises. This is partially because large language models aren’t improving at the breakneck pace of 2023 and 2024. Still, it’s also a recognition that customers not only want more accuracy from AI chats and agents, but also a more tailored and company-aligned user experience. I think this is also why we are seeing data-centric companies like Salesforce and MongoDB investing more in AI features and acquisitions.
Stay tuned for more as I dive into these topics this fall.
OpenAI had a monster week. And kudos to its marketing team for taking advantage of the slow summer months to get all eyes on the company when competitors largely stay dormant. The biggest news was GPT-5; after a two-year wait, we now see a new model that is a big step forward from its predecessor. While it did post impressive bench scores, in some ways GPT 5 is catching up with and not outpacing Claude, Gemini, and even in a couple of cases Grok. But while we may not see earth-shattering performance, we do see a new level of richness and breadth from the model, and I expect it will once again be the general-purpose AI standard for some time.
The day before GPT-5 was announced, OpenAI also pushed out two new models as open source. This was a big step for the company, as open APIs and not models have been its mode of operation. But what may be more notable than the strategic pivot is that these models — despite being hyped as available on all the major clouds on day 1 — are targeting PCs and mobile devices. This will surely open up some new doors into edge use cases and possible hybrid AI scenarios. For example, a local model could be a constant context collector based upon user inputs and actions, which could then feed into larger cloud-based models for heavier reasoning exercises. Imagine a situation where your phone is able to adjust the settings for any car you get into, even if you have never sat in it before. So, while the clouds will sell instances with these smaller models, I am way more interested to see what happens outside the cloud with these new models (as well as with Llama and others that compete with it).
Despite all of this progress with models, I do think that we are starting to see a plateau in terms of AI capabilities for enterprises. The missing piece will be the more effective use of corporate data to drive more effective context into models that are already pretty good. Lately, I have been turning my attention to DevOps and data pipelines, where I think the next AI breakthroughs will happen.
Last week, I had the chance to get briefed by Snowflake about its Cortex AI capabilities, and I learned quite a bit. While the ability for AI to generate SQL data queries is a key feature for business analysts, I was just as impressed with how Cortex AI was able to present outputs in text, tabular, and graphical formats. It also helps that Snowflake is data-agnostic and cross-cloud.
Cortex AI reinforces my belief that we are approaching a new phase in the evolution of enterprise AI. Even though experts have been saying that great data is the foundation for great AI since the advent of large language models years ago, we are only recently seeing that notion take hold in enterprises. This is partially because large language models aren’t improving at the breakneck pace of 2023 and 2024. Still, it’s also a recognition that customers not only want more accuracy from AI chats and agents, but also a more tailored and company-aligned user experience. I think this is also why we are seeing data-centric companies like Salesforce and MongoDB investing more in AI features and acquisitions.
Stay tuned for more as I dive into these topics this fall.
Broadcom announced the launch of its Jericho4 Ethernet switch silicon, supporting the scale-out requirements of AI across datacenters. With up to 51.2 Tbps of lossless Ethernet, Jericho4 supports scale to over 1 million XPUs as they span racks, clusters, and even sites. Built on a 3nm process node, it brings considerable performance gains, with Broadcom claiming a 40% power reduction per bit. It is set to deliver that 51.2Tbps of deep-buffered capacity across 36,000 HyperPort interfaces, each at 3.2Tbps bandwidth.
Broadcom owns about 80% of the Ethernet chip market. And Jericho4 is an example of why the company dominates. Despite its market position, the company keeps its foot on the gas to make sure its chips are the design point for connecting the datacenters of tomorrow. And it does this against a competitive NVIDIA that looks to keep pushing its Spectrum-X platform as a vertically integrated networking stack.
AMD announced its quarterly earnings, and I must say I’m a bit surprised by the datacenter numbers. Its $3.24 billion in revenue represents year-over-year growth of 14%. This would normally be considered a good number; but when the company has posted 61%, 69%, 122%, and 115% YoY growth in the four previous quarters, the number seems small. And I am especially surprised after seeing such strong numbers from TSMC in its earnings a couple of weeks back, and Intel’s $3.9 billion in DCAI revenue (nearly exclusively Xeon) in its most recent earnings.
I’m not a financial guy, but it’s hard to get my head around the China export restrictions for the MI308 as being the only driver for an operating income number that went from $743 million to negative-$155 million, and an operating margin that went from 26% to negative-5%, year over year. All on that 14% growth in revenue.
I wonder if, in addition to the impact of export restrictions on the MI308, we’re seeing a slowdown in the company’s GPU shipments overall — silicon that was delivering high margins. That would mean what we see is a bigger contribution from EPYC, which is being priced very competitively against Intel’s Xeon (and homegrown Arm silicon) in the hyperscale market.
EPYC has had a strong showing in the hyperscale market. AMD has to continue to drive that transactional commercial business to fully take advantage of EPYC’s momentum. The company has been talking about the enterprise segment being slow to move for some time — and it is. However, after seven years in-market, I would expect this segment to have moved past its inertia.
EPYC has a strong value proposition and market position, and has for some time. We should have already seen accelerated growth — not simply growth in the hyperscalers supporting enterprise through SaaS, IaaS, and other cloud services. And it’s a little surprising to hear comments in its quarterly earnings about how the company is fully funding enterprise GTM — this should be a given.
Informatica’s Q2 2025 results show a company shifting from its legacy roots toward a high-margin, cloud-first, and AI-driven model. Total revenue reached $407.3 million, up 1.7% year-over-year, with the real growth driver coming from cloud subscriptions, which climbed 30% to $209.9 million and now account for 73% of total subscription revenue. Cloud ARR hit $901 million, up 28%, helping lift total ARR to $1.72 billion. Operationally, Informatica broke even on a GAAP basis while delivering $109.4 million in non-GAAP operating income and maintaining a solid 26.9% margin. Free cash flow of $58.7 million reflects disciplined capital management. On the product side, the company introduced AI Agent Engineering within its Intelligent Data Management Cloud, a no-code tool slated to launch in fall 2025 that will enable enterprises to design and orchestrate multi-agent AI systems across AWS, Azure, Google Cloud, Snowflake, Salesforce, and other platforms.
These results reinforce the success of Informatica’s business model for transformation.The strong cloud ARR growth, combined with high retention rates, signals a scalable and predictable revenue stream that can fund continued innovation. Its recently released AI Agent Engineering adds another layer of differentiation, positioning the company as an orchestrator of multi-cloud and multi-agent AI environments — an area where enterprise demand is expected to accelerate. The growing share of revenue from cloud subscriptions suggests ongoing enhancements to the platform, while the new AI orchestration capability shows the promise to reduce integration complexity and speed time-to-value for AI initiatives. The pending deal for Salesforce to acquire Informatica still adds some strategic uncertainty, but if it materializes, it could deepen Informatica’s ecosystem reach and make its AI and cloud services even more integral to enterprise data strategies.
In manufacturing, AI failures usually start long before the model — they start with the data it’s fed. The real problem? Feeding AI a messy diet of unprepared data. Manufacturing plants are overflowing with sensor and IoT readings, but much of it is incomplete, inconsistent, or stuck in legacy systems. Without basics such as timestamps, equipment IDs, and standardized formats, even the smartest AI will make bad calls.
Getting data “AI-ready” isn’t glamorous — it’s process work. Clean it up, add the right metadata, and make sure it’s consistent everywhere. Lock it down with proper governance so the right people have access and the wrong ones don’t. Tear down the wall between OT and IT so data flows from the factory floor to the C-suite without translation headaches.
Old batch systems are outdated and should be retired. Modern manufacturing operates in real time, and so should your data. A data lakehouse setup offers the scale of a data lake with the structure and governance of a data warehouse, making it ideal for both deep analysis and quick insights. Combine this with a hybrid edge-to-cloud approach to promptly identify and address issues while the heavy AI training takes place in the cloud.
The manufacturers that win with AI in 2025 won’t just have clever algorithms — they’ll have data that’s clean, contextual, and ready to roll. That’s how you cut downtime, improve quality, and turn real-time insights into real-world results.
I had an opportunity to spend time with the Outshift by Cisco leadership team at Cisco Live U.S. 2025 to discuss its agentic AI AGNTCY framework. Recently, AGNTCY moved to the Linux Foundation, and this development is significant for two reasons. First, it has the potential to dramatically scale and accelerate the development and adoption of AGNTCY’s agentic framework. As a Linux Foundation project facilitated through working groups, AGNTCY will further refine agent discovery, identity, messaging, and observability. These core features are the right ones to focus on. One of the biggest challenges in realizing the full potential of agentic workflows will lie in identity access management and orchestration.
Second, with AGNTCY’s graduation to the Linux Foundation, new members including Dell Technologies, Google Cloud, Oracle Cloud Infrastructure, and Red Hat now join it. From my perspective, this validates the work that Outshift has been doing, plus it marshals formidable resources to ensure AGNTCY’s success.
J.P. Morgan Payments launched its Supply Chain Finance (SCF) solution in Oracle Fusion Cloud ERP. In the initial launch with FedEx, FedEx suppliers can opt for early payment based on FedEx’s superior credit rating, improving liquidity while optimizing working capital. The integration through Oracle B2B reduces the time to deployment from months to a simple setup process within the ERP. The platform also supports integrated banking, AI-driven expense processing, and virtual cards.
PNC Bank is following a similar path, integrating its PINACLE Connect treasury platform into Oracle Fusion Cloud ERP. This gives corporate clients direct access to banking functions such as balance retrieval, payments, approvals, and reconciliations without leaving the ERP environment. This aims to reduce process fragmentation, minimize manual errors, and speed financial operations.
These announcements show a shift to “invisible finance,” where banking and payment workflows are integrated into the enterprise. FedEx, PNC, and their clients benefit from this transformation thanks to quicker liquidity decisions, fewer reconciliation cycles, and fewer system handoffs, along with lower operational costs.
These improvements also allow Oracle to position itself as a stronger partner in embedded finance, making its ERP not just a system of record but also a system for cash flow execution. Ensuring high security standards and broad adoption among suppliers and treasury teams is crucial. Done right, these kinds of embedded models could redefine how enterprises think about ERP and financial operations over the next three to five years.
SYSPRO, which provides ERP software for manufacturers and distributors, has acquired DataScope Systems — a warehouse management system (WMS) provider that has been closely tied to Syspro’s platform for more than 20 years. This is Syspro’s third acquisition in a year, following Nexsys and riteSO, and comes under the ownership of Advent International, the private equity company that took a majority stake in 2024.
By bringing Datascope in-house, SYSPRO can now offer customers a unified ERP and WMS solution. For customers, that means better inventory visibility, tighter control over supply chain operations, and improved efficiency — without the disruption often seen when trying to integrate two separate systems. Because the products have been technically aligned for decades, the transition should hopefully be relatively smooth.
This deal strengthens SYSPRO’s competitive position. Solutions partners should have the ability to provide a more complete, integrated package to customers, making it easier to sell into more complex warehouse and distribution environments. It also gives SYSPRO the ability to speed up product innovation and expand globally, in line with Advent’s growth strategy.
I see this as a timely move. Instead of spending years building a WMS from scratch, SYSPRO gains a proven product with an existing customer base. The risks will be in integrating the teams, ensuring customers see value, and competing with ERP rivals such as Oracle, SAP, Infor, Acumatica, and Microsoft.
Battlefield 6’s first early access window and open beta was a successful one, with more than 500,000 playing at the peak. This was also an effective test of EA’s ability to scale servers based on demand, and to work through player feedback and implement fixes on the fly. I think that right now EA is just looking to figure out how players feel about the changes made after the Twitch streamers played and after this first Open Beta weekend. One thing I did notice is that while EA has its own EA App, the vast majority of my gaming friends were playing on Steam. So much so that I actually uninstalled the EA App version of the game and reinstalled the Steam version so that it would be easier to play with friends — which is ultimately the best experience in any Battlefield game. EA has said that this open beta is the biggest in the company’s history.
Last week, OpenAI and AWS announced availability of the open-weight gpt-oss-20b and gpt-oss-120b models, making these LLMs available beyond the Azure ecosystem. AWS customers can access these models in Amazon Bedrock and Amazon SageMaker JumpStart. In parallel, Microsoft announced gpt-oss-20b for Azure AI Foundry and Windows AI Foundry.
The gpt-oss-20b model is particularly interesting for edge AI because it is small enough to run on appropriately scaled embedded platforms and should perform nicely on industrial edge servers. Mixture-of-experts (MoE) sparse activation can reduce the total parameters to less than 4 billion per token, enabling inference on embedded platforms with as little as 16GB of memory. At first glance, you might not think this open model is a game-changer for edge AI because open-weight models have been available for a long time. However, this announcement is significant because of the models’ performance, familiarity, maturity, Apache 2.0 license, compatibility with AWS infrastructure, and MoE capabilities.
LLMs in small edge platforms enable a new wave of compelling use cases for industrial and consumer embedded applications. For instance, imagine using conversational UI in your car with no cloud dependencies: “Hey, car, let’s go to my daughter’s condo in Port Aransas. Take the 130, and stop 20 minutes for lunch at the Whataburger in Beeville. So, when will we arrive?” No keywords to remember and no rigid syntax, just plain-language, results-oriented prompts. Beyond this conversational use case, local inference with mainstream LLMs enables high-quality perceptive, prescriptive, agentic, and orchestrated modalities without cloud dependencies, privacy leakage, or need for continuous network access.
My only concern about these announcements is how quickly OpenAI will release new model versions. Amplifying this concern, Microsoft announced the availability of OpenAI’s GPT-5 in consumer, developer, and enterprise offerings two days after the open-weight announcement. We’ll see how this plays out, but I remain optimistic. Given the extremely positive industry reaction, I do not expect OpenAI to back away from the open-weight commitment.
The lawsuit brought by a California man against Microsoft for discontinuing Windows 10 support is a bit perplexing to me, especially considering that Windows 10 has been around for a decade now, and for a while you could upgrade to Windows 11 for free. I don’t think it’s fair and reasonable to expect Microsoft — or any software company — to support any piece of software beyond 10 years without an explicit agreement to do so. Windows 10 itself is not even a piece of enterprise piece software in this instance, plus Microsoft offers a remedy to continue to support users for $30, so it’s not like the company has completely cut off users who want to stay on Windows 10.
QuamCore is an Israeli startup that recently secured $26 million in Series A funding to develop a low-power superconducting trapped-ion quantum computer purported to have the potential of scaling to 1 million qubits within a single cryostat. QuamCore’s funding was led by Sentinel Global and other investors, plus a non-dilutive grant from the Israel Innovation Authority. QuamCore’s goal of a million error-corrected qubits, if it can be met, would enable the achievement of elusive real-world applications in areas such as drug discovery, advanced materials, and complex logistics systems.
The funding will allow QuamCore to eliminate a multigenerational product roadmap and instead move directly from prototype to a first-generation processor. Part of the funding will enable QuamCore to build a dedicated quantum lab to speed up its chip production and prototype integration.
QuamCore attributes its breakthrough to physically relocating its ultra-low-power superconducting control logic from an exterior position to inside the cryostat. A cryostat is a specialized refrigeration system that provides extremely cold temperatures near absolute zero. A cryogenic environment like this is needed for the creation and operation of superconducting qubits. Qubits are sensitive to environmental noise and thermal energy, which cause them to lose their quantum state and cause computational errors. Placing qubits inside a cryostat minimizes errors and helps them remain in a stable quantum state for a longer period.
Locating logic circuits inside the cryostat also eliminates much of the external cabling and thermal barriers that previously limited scalability. QuamCore’s architecture also incorporates built-in error correction, which makes fault tolerance possible. No date was given for when a prototype might be available.
Quantum Art has built a full-stack quantum computer through the use of multi-qubit gates capable of implementing 1,000 standard two-qubit gates in a single operation. This is made possible by simultaneous multi-tone, multi-mode coherent control over all qubits on a reconfigurable quantum processing unit (QPU). According to its most recent press release, Quantum Art has demonstrated a stable, 200-ion chain with ions spaced five microns apart.
Keeping ion chains 50 to 75 ions long is what I consider to be the upper limit because precise ion control is difficult on the outer ends of the chain. To me, 200 ions long is an eye-popping length. Even though this is one of the longest linear ion chains in a commercial-grade quantum computer, Quantum Art is still able to maintain massive entanglement and maintain full connectivity and stability for all the ions. It is also critically important to maintain linearity and stability for long ion chains. Quantum Art scientists accomplish this through precise trap geometry and electrode fabrication, ultra-low heating rates, and a low-noise cryogenic environment.
Quantum Art’s innovation opens the door to future systems with 1,000-ion registers. These larger systems will be composed of modular, optically segmented cores that can operate in parallel, offering high qubit connectivity and dynamic reconfigurability.
Quantum Art is currently working on two key projects. First, it is preparing for the initial release of its commercial 50-qubit Montage system. It is also developing a 1,000-qubit quantum processing unit called the Perspective platform that it says will be able to demonstrate a commercial quantum advantage by 2027.
While Quantum Art is focused on delivering a 50-qubit system soon and a 1,000-qubit platform by 2027, another company, IonQ, uses a trapped-ion architecture that targets significantly more powerful systems in that same period. By 2027, IonQ aims to have a system with 10,000 physical qubits to create 800 logical qubits. This is a considerable jump from Quantum Art’s goal of 1,000 physical qubits in the same timeframe.
Logical qubits are a key component for fault-tolerant quantum computing, as they represent error-corrected qubits that are much more powerful than noisy physical qubits. IonQ is focused not only on scaling the number of physical qubits but also on building systems with high-fidelity, error-corrected operations. IonQ also does not require cryogenics. It has a big advantage because it operates at room temperature, which expands the range of applications and business segments it can serve.
Riverlane has released a preview of its 2025 Quantum Error Correction (QEC) survey findings, and — as with the 2024 edition — the findings contain some interesting and surprising data. With offices in Cambridge, England, and Boston, Massachusetts, and more than 100 employees, Riverlane researches and develops quantum error correction, quantum computing, and industry applications.
Based on its 2025 survey, Riverlane concluded that companies using quantum need to bridge the gap between awareness of QEC and having the readiness and training necessary to use it. The survey found that 95% of quantum professionals consider QEC essential for scaling quantum computing; but it also found that a lack of training and knowledge, a lack of best practices, and insufficient access to large, stable quantum computers for experiments are preventing the knowledgeable use of QEC.
Because there is a large and immediate requirement for QEC, the report recommends that companies should treat it as a priority. Given that the survey indicates most companies rely on external resources to implement QEC, it recommends that they should (a) form partnerships with companies that have QEC expertise, (b) invest in creating internal QEC, and (c) make roadmap adjustments to accommodate hardware that integrates QEC into it.
Full-service companies like IBM can offer significant benefits by providing a complete ecosystem for quantum computing. IBM also offers integrated solutions, as well as technical support that includes experts for guidance on implementing QEC and developing fault-tolerant algorithms. IBM also offers a range of quantum computers and pricing tiers for QEC research and implementation.
Bad actors continue to improve the sophistication of cyberattacks, and reactive security tools used during this era of increased cyber threats are simply not enough to ensure the highest levels of protection. DNS continues to prove its value in this regard by providing preemptive security capabilities through the identification of potentially malicious domains that could become weaponized and used for illicit gain.
At Black Hat USA 2025, Infoblox used the event to showcase its DNS-based threat defense platform and several enhancements, including Protection Before Impact, Security Workspace, Detection Mode, and Asset Insights integrations. From my perspective, these capabilities provide organizations with a fortified and proactive security posture toolset to address today’s dynamic threat landscape. This Moor Insights & Strategy research brief dives deeper into the announcement and includes an example of a customer that achieved an optimized set of outcomes as a result.
Globant’s “AI or Die” framework asserts that for sports organizations, adopting AI is a strategic necessity, not an option. The company argues that AI must be integrated across core functions to maintain a competitive edge. This includes using AI to optimize athlete performance, enhance operational efficiency, and create new revenue streams through fan engagement. The central premise is that organizations failing to embrace AI risk becoming irrelevant in a rapidly evolving, data-driven sports industry. The company is clearly using this as a marketing message to sell its services; regardless, turning “reinvent yourself” into “AI or Die” certainly caught our attention. Robert and I talked about this on last week’s Game Time Tech podcast, alongside a Wimbledon recap and discussion of the Premier League’s new deals with Microsoft and Adobe. Check out that episode to catch our insights on these things and more.
AT&T has expanded its Office@Hand business communications portfolio through a partnership with RingCentral. The new offerings include Office@Hand Contact Center, powered by RingCX, and the conversational intelligence solution Office@Hand RingSense. The Contact Center platform uses AI to manage interactions across more than 20 channels, providing a unified view of the customer journey and offering automated transcripts and coaching insights for human agents. RingSense supports sales and support team performance with post-call analysis features like automatic summaries and sentiment analysis. These solutions are now available to all customers, building on the existing partnership with AT&T.
This partnership was a key highlight in RingCentral’s strong Q2 2025, which saw 6% year-over-year subscription growth and a 7% increase in ARR. The company posted its first-ever positive GAAP operating and net income, indicating operational improvement. Besides AT&T’s launch of RingSense and RingCX, the company extended its relationship with NiCE. RingCentral’s AI solutions continue to gain traction, with more customers adopting RingSense, RingCX, and the AI Receptionist. RingCentral also generated stronger free cash flow, paid down debt, expanded its share buyback, and raised its outlook for the year. All in all, the company’s focus on translating AI technology into practical, customer-driven outcomes has supported stronger engagement and performance this quarter.
Last week 8×8 posted its first quarter of revenue growth in more than two years as its Q1 FY 2026 earnings reached the high end of guidance at $181.4 million. This was driven by a 2% increase in service revenue YoY and a 12% increase in usage-based revenue. The company has also reduced its debt by 40% since mid-2022 and expanded its voice integrations for Microsoft Teams by 30%. However, these positive trends are accompanied by ongoing challenges. Q1 GAAP earnings missed expectations, with EPS at $0.08, as rising costs associated with expanding usage-based services challenged margins. While new consumption-driven offerings are improving the top line, legacy revenue continues to decline. Achieving sustainable, profitable growth remains a key challenge. The company has shown an ability to maintain margin discipline, but sustaining this through the transition to usage-based offerings is a key challenge as competitive pressures persist. Ultimately, while 8×8 made clear progress getting back to growth and reducing debt, its test now is to keep margins steady and prove that its shift to usage-based offerings can be a sustainable business model.
Celona recently announced its new AerFlex private 5G cellular networking architecture, which could raise its market competitiveness by reducing the friction tied to deployments in both operational technology (OT) and information technology environments.
As with Wi-Fi, a customer implementing this solution can simply hang access points and be done — no need to deploy physical onsite infrastructure thanks to a cloudified core. Secure local breakout also ensures that data is safeguarded — an important consideration given the growing use of data from deployments like these within large and small language models to feed modern AI applications. The potential capital and operational expense savings could be significant, avoiding incremental infrastructure investments and dramatically reducing the need for dedicated engineering specialists in branch locations.
Forbes Articles Published
- Qualcomm Execs Discuss Plans for Growth in Automotive, IoT and Beyond (Patrick Moorhead)
- Samsung Galaxy Tab S10 Ultra — The Ultimate Samsung Tablet? (Anshel Sag)
Research Papers Published
- Infoblox – Preemptive Security Leveraging the Power of DNS (Will Townsend)
Research Notes Published
- BROADCAST ANALYSIS: Patrick Moorhead Discusses Intel on CNBS Money Movers, August 7, 2025 (Patrick Moorhead)
- RESEARCH NOTE: Arm Zena CSS Enables Accelerated Automotive Development (Anshel Sag)
- BROADCAST ANALYSIS: Patrick Moorhead Discusses Apple Earnings on CNBC Closing Bell Overtime, July 31, 2025 (Patrick Moorhead)
- RESEARCH NOTE: Insta360 Link 2 — The True Power of AI in Webcams (Anshel Sag)
Analyst Insights Published
- RESEARCH NOTE: Pixel 9a Review — Exceptional Value, Mediocre Camera (Anshel Sag)
- AI Will Make the ‘Infinite Workday’ Worse — Unless We Act Differently (Melody Brue, Patrick Moorhead)
- AWE 2025 Recap — Is XR Going Mainstream Yet? (Anshel Sag, Patrick Moorhead)
- MI&S Weekly Analyst Insights — Week Ending August 1, 2025 (MI&S Analyst Team)
- Allowing the NVIDIA H20 Into China is a Good Start — But Not Enough (Patrick Moorhead)
Podcasts Published
The Enterprise Applications Podcast (Melody Brue, Robert Kramer)
Game Time Tech (Melody Brue, Robert Kramer)
G2 on 5G (Will Townsend, Anshel Sag)
Don’t miss future MI&S Podcast episodes! Subscribe to our YouTube Channel here.
Press Citations
Intel / Trump on Intel CEO / Patrick Moorhead / BBC
Trump calls for Intel boss to resign immediately, alleging China ties
Intel / Trump on Intel CEO / Patrick Moorhead / Fierce Networks
Op-ed: Trump’s attack on Intel is a big dumb distraction
Intel / Trump on Intel CEO / Patrick Moorhead / Fierce Networks
Tan sees Intel board support amid call by Trump to resign over China ties
Intel / Trump on Intel CEO / Patrick Moorhead / Investors Observer
Intel’s $8B CHIPS Act grant at risk after Trump Blasts CEO’s China ties
Intel / Trump on Intel CEO / Anshel Sag / Business Today
Why is Donald Trump pushing Intel CEO Lip-Bu Tan to resign?
Intel / Trump on Intel CEO / Anshel Sag / Mobile World
Trump calls for Intel CEO to resign
Intel / Trump on Intel CEO / Anshel Sag / The Columbus Dispatch
President Donald Trump, Sen. Bernie Moreno call on Intel CEO to resign, citing China ties
Intel / Trump on Intel CEO / Anshel Sag / The Times
Trump demands ‘conflicted’ Intel chief quits over China links
Intel / Trump on Intel CEO / Anshel Sag / Reuters
Reactions to Trump demanding Intel CEO’s resignation
Intel / Trump on Intel CEO / Anshel Sag / Tahawultech
Trump demands immediate resignation of Intel CEO
TV/Broadcast Appearances
Trump & Intel / Patrick Moorhead / CNBC Money Movers
If we lose Intel as a foundry it’s a big national security risk, says Moor Insights’ Patrick Moorhead
AMD Earnings / Patrick Moorhead / Prof G Podcast
Succession: Jerome Powell Edition | Prof G Markets
New Gear or Software We Are Using and Testing (New)
- Perplexity Comet Browser (Jason Andersen)
- Alienware 18 Area 51 Gaming Laptop (Anshel Sag)
- Kiro IDE (Jason Andersen)
- IBM Concert (Jason Andersen)
- Samsung Galaxy Fold7 (Anshel Sag)
- XREAL One Pro (Anshel Sag)
- XREAL Eye AR Glasses (Anshel Sag)
- AWS Q CLI (Jason Andersen)
- Surface 12” laptop (Anshel Sag)
- Anonymous Tablet (Anshel Sag)
- OMEN MAX 16 Gaming Laptop (Anshel Sag)
- Alienware Area 51 Desktop & 27” 4K OLED monitor (Anshel Sag)
- Samsung Galaxy Tab S10 Ultra (Anshel Sag)
- RTX 5060 Ti (Anshel Sag)
- Elgato Wave DX Microphone (Melody Brue)
- Elgato Wave Desk Stand & Master Mount (Melody Brue)
- Poco F7 Ultra Smartphone (Anshel Sag)
- 2.0 Antec Flux Pro Case (Anshel Sag)
- 2.1 GeForce RTX 5070 and 5070 Ti (Anshel Sag)
- Eve Systems dimmer switch – Matter-based and Thread-connected (Bill Curtis)
- Sony CRE-C20 OTC Hearing Aids (Anshel Sag)
- Steelseries Arctis Nova Pro Wireless Headset (Anshel Sag)
- HP EliteBook X G1a (Anshel Sag)
- Kindle Colorsoft (Anshel Sag)
- XREAL One AR Glasses (Anshel Sag)
- Various Matter devices (Bill Curtis)
Events MI&S Plans on Attending, in Person or Virtually (New)
Unless otherwise noted, our analysts will be attending the following events in person.
- HP Summer Gaming Event, August 18, Los Angeles (Anshel Sag)
- Hot Chips, August 25-26 (Matt Kimball – virtual)
- VMware Explore, August 25-28, Las Vegas (Matt Kimball)
- IBM US Open & Innovation Studio, August 20-12, New York (Melody Brue)
- T-Mobile Technology Innovation Summit 2025, August 27-28, Bellevue, Washington (Patrick Moorhead, Melody Brue, Anshel Sag, Will Townsend)
- HP Summer Gaming Event, August 18, Los Angeles (Anshel Sag)
- Hot Chips, August 25-26 (Matt Kimball – virtual)
- VMware Explore, August 25-28, Las Vegas (Matt Kimball)
- IBM US Open & Innovation Studio, August 20-12, New York (Melody Brue)
- T-Mobile Technology Innovation Summit 2025, August 27-28, Bellevue, Washington (Patrick Moorhead, Melody Brue, Anshel Sag, Will Townsend)
- JFrog Swamp Up, September 6-8, Napa (Jason Andersen)
- Splunk .conf25, September 8-10, Boston (Will Townsend)
- Box Analyst Summit, September 11, San Francisco (Melody Brue)
- Zoomtopia, September 17 (Melody Brue – virtual)
- MongoDB local, September 17, New York (Jason Andersen)
- Okta Oktane, September 23-25, Las Vegas (Will Townsend – virtual)
- Cloudera EVOLVE25, September 24-25, New York (Robert Kramer)
- Connected Britain, September 24-25, London (Will Townsend)
- HashiConf, September 24-25, San Francisco (Jason Andersen)
- Webex One, September 28 – October 1, San Diego (Melody Brue)
- SailPoint Navigate, September 29 – October 2, Austin (Will Townsend)
- Intel ITT, September 28 – October 2, Phoenix (Matt Kimball)
- IBM TechXchange, October 6-9, Orlando (Jason Andersen, Matt Kimball, Melody Brue)
- MWC America, October 8-10, Las Vegas (Will Townsend – virtual)
- Oracle CloudWorld, October 13-16, Las Vegas (Matt Kimball, Melody Brue)
- Supercomputing ‘25, November 16-21, St. Louis (Matt Kimball)
- Lenovo Global Industry Analyst Conference 2025 + Tech World, October 20-23, Raleigh, North Carolina (Melody Brue, Matt Kimball)
- Techritory, October 22-23, Riga (Will Townsend)
- Infor Analyst Summit, October 28-30, Orlando (Robert Kramer)
- Adobe MAX, October 28-30, Los Angeles (Melody Brue)
- Cisco Partner Summit, November 3-5, San Diego (Will Townsend)
- Smartsheet ENGAGE, November 4-7, Seattle (Melody Brue)
- Commvault SHIFT, November 11-13, New York (Robert Kramer)
- Dell, Analyst Summit, dates TBD, Austin (Matt Kimball)
- Microsoft Ignite, November 17-21, San Francisco (Robert Kramer)
- NTT R&D Forum, November 19-21, Tokyo (Will Townsend)
- AWS re:Invent, December 1-5, Las Vegas (Patrick Moorhead, Robert Kramer, Matt Kimball, Will Townsend, Jason Andersen)
- Cisco Analyst Summit, San Diego (Matt Kimball)
- ZohoDay 2026, February 16-19, Austin (Melody Brue)
- Nutanix .NEXT ‘26, April 2-9, Chicago (Matt Kimball)
Subscribe
Want to talk to the team? Get in touch here!








