Upscale AI Brief
Other·AI Networking·Series A-1 (extension)·🇺🇸
Summary
Recently raised· Series A-1 (extension) this monthUpscale AI is a full‑stack, purpose‑built AI networking vendor attacking the AI datacenter interconnect gap with a software‑centric stack (SkyHammer™ architecture, Open Ethernet Fabrics with NVIDIA Spectrum‑X + SONiC). The company sits as a niche challenger to incumbents NVIDIA, Broadcom and Cisco — offering tighter system integration for large‑scale training/inference rather than raw silicon scale. Momentum signals are real: multiple large, dated funding rounds and a disclosed strategic NVIDIA collaboration, plus investors that include Premji Invest, Intel Capital, Mayfield and Qualcomm Ventures. Named customers (ESUN, UALink, UltraEthernet) and an enterprise‑custom pricing model point to an enterprise route‑to‑market, but the most important tension is uncertain commercial scale (no public ACV or verified customer references) versus heavy competition from deeply entrenched silicon and switch vendors.
Bull case
- Founding team has repeat exits and deep networking‑silicon credibility (co‑founders Barun Kar and Rajiv Khemani supported by experienced ASIC/systems hires).
- Strategic technical signals: public SkyHammer™ architecture and announced Open Ethernet Fabrics (NVIDIA Spectrum‑X + SONiC) demonstrate product focus on AI training/inference topologies.
- Strong institutional backers on the cap table (Premji Invest, Intel Capital, Mayfield, Qualcomm Ventures, Salesforce Ventures) and multiple large funding events indicate investor conviction and access to follow‑on capital.
- Named customers on the site (ESUN, UALink, UltraEthernet, OCP SAI/ESUN, SONIC) provide early commercial validation and suggest traction with infrastructure buyers.
Watch-outs
- Competitive intensity: incumbents NVIDIA, Broadcom and Cisco dominate datacenter networking and have scale advantages in silicon, OEM relationships and go‑to‑market penetration.
- Commercial opacity: the company has not disclosed per‑customer ACV, verified customer reviews are absent, and the site does not state geographic go‑to‑market scope, making revenue runway and sales efficiency hard to assess.
- Funding/metric inconsistencies in public materials — site lists ‘raised over $100M’ while other reporting documents multiple large rounds (including a $100M seed, $200M Series A and $190M A‑1) and differing total funding figures — clouding runway and valuation clarity.
- Customer sentiment and third‑party references are thin: no credible public customer reviews were found that unambiguously refer to this Upscale AI, increasing diligence burden on deployment reliability and support.
From their website
Stats they publish
Pricing
Not publicly disclosed. The company website does not list its pricing on any of the 6 pages we read.
Product suite
- SkyHammer
Scale-up AI networking architecture that delivers deterministic latency, extreme bandwidth and predictable rack-scale performance
- Open Ethernet Fabrics
Open, interoperable Ethernet systems for heterogeneous AI clusters (NVIDIA Spectrum‑X + SONiC)
Target segments
Testimonials
“Upscale AI is entering at exactly the right moment with open, high-performance solutions built to scale.”
Crunchbase
verified profile
- Upscale AI Raises $190 Million Series A-1 at a $2 Billion Valuationstartupresearcher.com · 2026-06-22
- Upscale AI Raises $190M to Expand AI Networking InfrastructureVentureburn · 2026-06-22
- Upscale AI Raises $190 in Series A-1 FundingFinSMEs · 2026-06-22
- Upscale AI valued at $2b after funding extension; Premji Invest leads roundDealStreetAsia · 2026-06-22
- Upscale Raises $190 Million for AI Networking TechPYMNTS.com · 2026-06-22
- Upscale AI Adds $190 Million in Extension to Series A, Reaching Half-Billion Dollars in Total FundingBusiness Wire · 2026-06-22
- 拆解25笔超5000万A轮融资，AI最聪明的钱正在押注这三个方向ChinaVenture · 2026-06-09
- Cybersecurity Funding Holds Up At Robust LevelsCrunchbase News · 2026-04-20
Team
Upscale AI publicly lists a network‑centric founding pair (Barun Kar and Rajiv Khemani) with multiple repeat exits in networking silicon, supported by experienced hires in ASIC engineering, systems, and operations and several advisors/board members from the infrastructure ecosystem.
The named founders (Barun Kar and Rajiv Khemani) have deep prior experience building networking silicon and systems (co‑founding Innovium and earlier roles at Cavium and Palo Alto Networks), directly aligned with AI‑networking hardware and software.
Strength: Founding team with direct domain experience in networking silicon and systems (Innovium, Cavium, Palo Alto Networks) and a broad leadership bench across ASIC, systems engineering and operations. Gap: public record does not show a named CFO/finance lead.
- Barun KarCo‑Founder, CEO, President, Board MemberBarun KarCo‑Founder, CEO, President, Board MemberPalo Alto Networks, Innovium, CaviumCo‑founded Cavium (acquired by Marvell), Co‑founded Innovium (acquired by Marvell, 2021)
Serial networking entrepreneur who co‑founded and led multiple infrastructure companies and has repeatedly built silicon + systems stacks for networking.
Repeat founder - Rajiv KhemaniCo‑Founder, Executive Chairman, Board MemberRajiv KhemaniCo‑Founder, Executive Chairman, Board MemberInnovium, AuradineCo‑founded Innovium (acquired by Marvell, 2021)Repeat founder
Leadership
- Puneet AgarwalChief Technology OfficerPuneet AgarwalChief Technology OfficerVice President and CTO of Data Center, Marvell · InnoviumCo‑founded Innovium (acquired by Marvell, 2021)Repeat founder
- Jason LedgerwoodSenior Vice President of Systems Engineering & OperationsJason LedgerwoodSenior Vice President of Systems Engineering & OperationsCisco, Brocade, Flex, Palo Alto Networks
Over 25 years in procurement, manufacturing and operations across major networking and systems companies.
Ex top-tier - Sharada YeluriVice President of ASICSharada YeluriVice President of ASICAssociate Vice President of Engineering, Scale‑up Fabrics, Astera Labs · Senior Director of Engineering, Juniper NetworksEx top-tier
- Mohsen MoazamiSenior Advisor
- KK VermaVice President of Customer SupportKK VermaVice President of Customer Support
Customer support executive responsible for post‑sales support and customer success operations for networking and infrastructure products.
Market Size
A focused, published top-down figure for the 'AI in Networks' market (the closest published category for hardware+software optimized for large-scale AI training/inference) is USD 8.67B (2023) per Grand View Research; other reputable reports present higher 2024–2026 baselines but broadly agree the AI-networking category is a multi-billion-dollar, fast-growing global market. Upscale AI sells full‑stack AI networking hardware and software into this global market, but the company has not published per-customer ACV or customer counts publicly, so a grounded bottom-up SAM and a defensible SOM cannot be derived from disclosed figures.
- Grand View Research ↗USD 8.67 billion in 2023; projected to reach USD 60.60 billion by 2030, CAGR 32.5% (2024–2030) · Global · 32.5% CAGR
- MarketsandMarkets ↗USD 10.93 billion (2024) → USD 46.87 billion (2029), CAGR 33.8% (2024–2029) · Global · 33.8% CAGR
- Intel Market Research ↗USD 14.3 billion (2026) → USD 45.7 billion (2034), CAGR 16.8% (2026–2034) · Global · 16.8% CAGR
- Company pricing modelenterprise_custom (no per-customer ACV published)Upscale AI site (known facts)
- Company traction / scale relevant to SOMRaised > $100M (seed) and subsequent reporting of $200M Series A in Jan 2026; valuation > $1B; no public ARR or customer counts disclosedUpscale AI press release and Bloomberg (known facts / web search)
- TAM chosen is the published 'AI in Networks' / 'AI in Networks Market' global figure (Grand View Research 2023). Several other reputable vendors publish differing top-down totals and growth rates (see MarketsandMarkets and Intel Market Research); the available top-down estimates vary materially by publisher and forecast horizon.
- Global region used throughout because Upscale AI targets hyperscalers/cloud/data-center operators globally and the available published market figures are global. If a country- or region-specific SAM is required, additional customer-count and pricing disclosures or region-specific market slices are needed.
Competitors
Top 3Incumbent silicon and switch vendors (NVIDIA, Broadcom, Cisco) dominate AI datacenter networking. Upscale AI is a niche challenger: full‑stack, AI‑optimized networking hardware+software focused on large‑scale training/inference.
- NVIDIAnvidia.com ↗Provides NVSwitch and high‑speed AI interconnects; Upscale offers purpose‑built full‑stack AI networking as a smaller, software‑centric alternative that complements and competes on system integration.
- Broadcombroadcom.com ↗Makes leading switch ASICs (Tomahawk family); Upscale differentiates with AI‑optimized hardware+software stack, while Broadcom wins on silicon scale and OEM penetration.
- Ciscocisco.com ↗Legacy networking incumbent with broad enterprise reach and switch portfolio; Upscale competes by specializing in large‑scale AI training/inference networks and faster, workload‑specific innovation.
Funding
- Seed Round$100MSep 17, 2025Maverick Silicon leadMayfield Fund leadCota CapitalStepStone GroupCelesta CapitalXora InnovationQualcomm VenturesMVP VenturesStanford UniversityTechCrunch ↗
- Series A$200MJan 21, 2026Premji Invest leadTiger Global Management leadXora Innovation leadMaverick SiliconQualcomm VenturesProsperity7 VenturesMayfield FundStepStone GroupIntel CapitalTracxn ↗
- Series A$190MPremji Invest leadTiger Global ManagementStepStone GroupNVIDIAProsperity7 VenturesSeligman VenturesMaverick SiliconTemasekMayfield FundSalesforce Venturestechstartups.com ↗
- Series A-1 (extension)$190MJun 22, 2026Premji Invest leadNvidiaSalesforce VenturesTemasekSeligman VenturesMayfieldTiger GlobalStepStoneMaverick SiliconProsperity7Fortune ↗
Public reporting documents a $100M seed (Sep 17, 2025) co-led by Mayfield and Maverick Silicon, a $200M Series A (Jan 21, 2026) led by Tiger Global Management, Premji Invest and Xora Innovation, and a $190M Series A-1 extension (June 22, 2026) led by Premji Invest that the company says brings total funding to $500M and values the company at $2B.
- Press releases and major outlets (Fortune, Business Wire, PR Newswire) report total funding of $500M; the three itemized rounds sum to $490M. No public source in the reviewed corpus explains the $10M gap.
- Crunchbase/Tracxn extracts list different totals (Tracxn shows $300M) or fewer rounds; those database figures conflict with the company and press announcements and suggest the public record is still being updated across trackers.
- Investor lists per round are compiled from press accounts and tracker summaries; some outlets list additional participating investors (e.g., Tiger Global, StepStone, Prosperity7, Intel Capital, Seligman, Nvidia) and different groupings across stories.
Momentum
- FundingJune 22, 2026Upscale AI raised $190 million in a Series A-1 extension, valuing the company at $2 billion and bringing total funding to $500M.Business Wire ↗
- PartnershipMar 11, 2026Upscale AI announced it will deliver an open, interoperable AI fabric with NVIDIA Spectrum‑X Ethernet and AI‑optimized systems and software.Upscale AI (press release)
- FundingJan 21, 2026Upscale AI closed a $200 million Series A round led by Tiger Global, Premji Invest and Xora Innovation.Tracxn ↗
- LaunchSep 17, 2025Upscale AI launched with over $100 million in seed funding.TechCrunch ↗
Multiple dated funding rounds and a strategic NVIDIA collaboration in the last 12–18 months indicate accelerating commercial and financing momentum for Upscale AI.
- Several funding reports are widely syndicated across outlets; some article phrasing is duplicated between press releases and media stories.
Customer Sentiment
No credible public customer reviews were found that clearly refer to Upscale AI (upscaleai.com), the purpose-built AI networking hardware/software company. Available third-party review snippets refer to other products and services with similar names (sales engagement tools, image-upscalers, apps), preventing an evidence-backed customer-sentiment assessment for this company.
Search results and review snippets returned reviews for multiple different "Upscale" or similarly named products (sales-engagement Upscale.ai / Upscale on G2, Upscale.media image upscaler, Upscale.pro, Upscayl, etc.).
No review text explicitly references Upscale AI's networking hardware or software (upscaleai.com) or any customer experiences with that product.
Because of name collisions and the absence of clear, product-specific reviews, public customer sentiment for Upscale AI (upscaleai.com) is not publicly disclosed.
AI Visibility
We ask ChatGPT WITH web search about the company and show its answers plus the sources it cited — the real AI-search experience. The key signal is share of voice: does the company come up when AI is asked about its space, not just when asked by name.
Models asked ChatGPT · web search
Accuracy of what AI said: Accurate — Direct answer matches the company's public profile: Upscale AI is an AI-networking vendor (rack-scale, low-latency networking) and lists the founders, funding, team size, and product names as on the company site.
Upscale AI is a company specializing in high-performance networking infrastructure tailored for artificial intelligence (AI) workloads. Their solutions are designed to address the unique demands of AI applications, offering ultra-low latency and scalable networking to enhance AI training and inference processes. (upscaleai.com)
Founders and Leadership: The company was founded in 2025 by Barun Kar, who serves as CEO, and Rajiv Khemani, the Executive Chairman. Both founders have extensive experience in the tech industry, having co-founded and led ventures such as Palo Alto Networks, Innovium, and Cavium. (upscaleai.com)
Company Size and Stage: As of April 2026, Upscale AI has grown to a team of 129 employees. The company has secured over $300 million in funding, achieving unicorn status. (upscaleai.com)
Notable Facts:
- Funding Milestones: In September 2025, Upscale AI announced its launch with over $100 million in seed funding. (upscaleai.com) By April 2026, the company had raised an additional $190 million, bringing its total funding to over $300 million. (upscaleai.com)
- Product Development: The company is developing products like SkyHammer™, an AI-native architecture designed to deliver deterministic, low-latency networking for synchronized AI workloads.
- Leadership Expansion: In April 2026, Upscale AI expanded its leadership team by appointing Puneet Agarwal as CTO, Jason Ledgerwood as SVP of Systems Engineering & Operations, Sharada Yeluri as VP of ASIC, and Mohsen Moazami as Senior Advisor. (upscaleai.com)
Upscale AI's focus on open-standard AI networking infrastructure positions it as a significant player in the evolving AI and networking landscape.
Here is a concise list of notable companies and products in the AI networking space:
- NVIDIA: Provides full-stack networking solutions tailored for AI data centers, including NVLink, Quantum InfiniBand, and Spectrum-X Ethernet platforms. (nvidia.com)
- Hewlett Packard Enterprise (HPE): Following its acquisition of Juniper Networks, HPE introduced new Juniper-branded switches like the QFX5140 and QFX5252, designed for AI workloads. (itpro.com)
- Cisco Systems: A leader in enterprise networking, offering AI-driven networking solutions and extensive LAN/WAN offerings. (unanswered.io)
- Arista Networks: Specializes in high-performance data center networking, providing solutions optimized for AI and cloud environments. (en.wikipedia.org)
- Marvell Technology: Partnered with NVIDIA to integrate into the NVLink Fusion ecosystem, enhancing AI networking capabilities. (tomshardware.com)
- Huawei Technologies: Developed AI-integrated networking solutions, including the RAN Intelligent Agent ecosystem, to boost network efficiency. (grandviewresearch.com)
- Nexthop AI: Launched scale-out and scale-across switches engineered for hyperscalers and cloud data centers, focusing on power efficiency and deployment speed. (businesswire.com)
These companies are at the forefront of integrating AI into networking solutions, driving advancements in performance, scalability, and efficiency.
In the rapidly evolving AI networking space, several companies are at the forefront, offering innovative products and solutions. Here are some notable organizations:
1. Hewlett Packard Enterprise (HPE): At HPE Discover 2026, HPE introduced new networking products tailored for AI workloads, emphasizing the foundational role of advanced networking in developing agentic AI systems. Following its 2025 acquisition of Juniper Networks, HPE unveiled Juniper-branded switches like the QFX5140 and QFX5252, designed for edge-based AI inferencing clusters and scalable AI infrastructure, respectively. (itpro.com)
2. NVIDIA: NVIDIA offers a comprehensive networking portfolio purpose-built for AI data centers. Their solutions include NVIDIA NVLink™ for scale-up, Quantum InfiniBand and Spectrum-X™ Ethernet for scale-out, and BlueField® DPUs for infrastructure services, all designed to deliver high-performance, scalable, and secure AI factories. (nvidia.com)
3. Cornelis Networks: Cornelis Networks provides the CN5000 Omni-Path®, an open, end-to-end, high-performance, natively lossless, and scalable networking solution. It's designed to maximize AI and HPC performance by offering faster model training completion times and lower inference latency compared to other solutions. (cornelisnetworks.com)
4. DriveNets: DriveNets offers the AI Networking Fabric, a full-stack Ethernet-based solution providing InfiniBand-level performance for AI back-end fabrics, front-end, and storage networks. It's designed to support any GPU, cluster size, and application, ensuring high-performance and scalability. (drivenets.com)
5. Arista Networks: Arista has introduced next-generation 1.6 Terabit products for AI fabrics, such as the 7060XE7 Series, expanding Etherlink for scale-out and scale-up applications. Their AI Fabric solutions are built to handle the demands of AI workloads, offering scalability and high performance. (arista.com)
6. Broadcom: Broadcom unveiled the Thor Ultra, the first 800G AI Ethernet Network Interface Card (NIC), designed specifically for AI-focused data centers. Utilizing PCIe Gen6 x16, the Thor Ultra dramatically boosts data throughput, making it ideal for environments with intense AI workloads and large-scale processing. (techradar.com)
7. Dell Technologies: Dell introduced the PowerRack, a comprehensive and integrated rack-scale system designed to streamline AI and high-performance computing deployments. The PowerRack integrates computing, networking, and storage into a unified, scalable system, eliminating the complexities involved in assembling disparate components. (itpro.com)
These companies are leading the charge in developing networking solutions that cater to the unique demands of AI workloads, ensuring high performance, scalability, and efficiency.
The assistant reliably describes Upscale AI when asked by name, but it did not surface the company unprompted as a notable player in the broader AI‑networking category or in recommendation answers. The company’s direct description appears consistent with its public materials and can be used cautiously for initial diligence.
- AI answers can vary run-to-run and different queries/searches may surface different companies or details.
- Public web coverage for recent or fast-growing startups can be thin; verify funding, headcount, and leadership details directly from the company or regulatory filings before relying on them for investment decisions.
Risks & Red Flags
No material public risk signals found. An absence of public concerns isn’t a clean bill of health — early or private companies may simply not generate coverage.
No material public risk signals were identified; an absence of public reports is not a clean bill of health for a private or early-stage company that may not attract press or regulatory visibility.
This review excludes automated domain 'trust score' results or generic site-safety checker outputs, which are not treated as evidence of real-world incidents.