Broadcom CEO Hock Tan says the company has a “line of sight” to more than $100 billion in AI-chip revenue in 2027. That is a management forecast for chip content—custom XPUs, switch chips, DSPs and related silicon—not a promise of $100 billion in Nvidia-style GPU sales, complete AI racks or guaranteed profit. Broadcom’s case rests on six large custom-silicon customers, multigigawatt deployment plans, networking and packaging expertise, and supply commitments extending through 2028. CRN’s account of the fiscal first-quarter 2026 discussion provides the company’s explanation.
What Broadcom’s $100 billion forecast actually covers
Tan said Broadcom expects AI-chip revenue to exceed $100 billion in 2027. The figure is separate from total corporate revenue and refers to silicon supplied into AI infrastructure, including:
- Customer-specific XPUs, or AI accelerators.
- Ethernet and other switch chips used to connect large accelerator clusters.
- Digital signal processors (DSPs).
- Related silicon content supplied with those systems.
It should not be read as $100 billion of Broadcom-recognized revenue from every part of an AI data center. A customer’s capital spending, the value of a complete rack, Broadcom’s chip content, bookings, recognized revenue and eventual profit are different measures. When asked about Anthropic, Tan did not separate chip revenue from rack revenue, leaving that boundary unresolved in the public discussion. The earnings-call transcript records that exchange.
“Line of sight” is management’s expression of expected visibility, not formal revenue guidance or independently verified demand. The exact fiscal-to-calendar recognition timing for 2027 should be checked against Broadcom’s eventual filings.
#1 Best Overall
- NVIDIA Volta GV100 Architecture — 4,608 CUDA Cores, 640 1st-Gen Tensor Cores delivering 14 TFLOPS FP32 and 112 TFLOPS deep learning performance for AI training, inference, HPC, and scientific computing workloads
- 32GB HBM2 ECC Memory — 900 GB/s Bandwidth — High-bandwidth memory on a 4096-bit bus with ECC error correction provides the memory capacity and throughput required for the largest AI models, simulations, and datasets
- PCIe 3.0 x16 Interface — 250W TDP — Standard PCIe Gen3 connectivity with passive cooling designed for enterprise rack server deployment in HPE ProLiant, Dell PowerEdge, and Supermicro platforms with adequate chassis airflow
- NVLink — Scale to 96GB Unified Memory — Connect two V100 GPUs via NVLink at 300 GB/s bi-directional bandwidth to scale GPU memory from 32GB to 96GB for larger AI training and HPC workloads
- Multi-Precision Computing — Supports FP64 (7 TFLOPS), FP32 (14 TFLOPS), FP16 (112 TFLOPS) and INT8 precision modes for flexible deployment across training, inference, and scientific simulation workloads
Why custom XPUs are the foundation
An XPU in this context is a custom accelerator designed around a particular customer’s workloads. Broadcom is not describing one standardized accelerator that any enterprise can order like a commercial GPU. Instead, it works with a small number of very large customers to define, design, package and produce silicon tailored to their systems.
When custom silicon makes economic sense
- Predictable scale: A hyperscaler expects enormous, repeatable workloads rather than occasional experimentation.
- Optimization: The design can target performance per watt, latency or total cost for a known workload.
- Engineering capacity: The customer can help specify architecture, software and validation.
- Amortization: Large volumes spread design and qualification costs across many units.
- Supplier diversification: A purpose-built accelerator can reduce reliance on a general-purpose accelerator platform for selected tasks.
Broadcom says its role spans silicon design and intellectual property, high-speed SerDes, networking, advanced packaging and high-volume manufacturing. Tan’s operational point was that a laboratory prototype is not enough: the harder test is producing roughly 100,000 chips quickly with acceptable yield and cost. That is a management claim, not an independently measured industry advantage, but it explains why Broadcom presents manufacturing execution as part of the product.
The six-customer engine
Broadcom says six major customers underpin the custom-silicon opportunity. Public coverage identifies four of them; the other two have not been named in the available reporting.
| Customer | Publicly described program | Scale discussed | What remains unknown |
|---|---|---|---|
| Continuing TPU program, with seventh-generation demand in 2026 and stronger later-generation demand expected | Not stated publicly in the cited coverage | Customer-level pricing, contract terms and Broadcom revenue | |
| Anthropic | TPU-based compute program | About 1 GW in 2026; more than 3 GW projected in 2027 | How much is chip content versus rack value; contractual economics |
| Meta | MTIA custom-accelerator roadmap | Multiple gigawatts projected in 2027 and beyond | Exact volumes, designs and supplier allocation |
| OpenAI | First-generation XPU deployment | More than 1 GW of compute capacity projected for 2027 | Commercial terms and Broadcom’s recognized revenue |
| Customer four | Broadcom said shipments were strong | Shipments expected to more than double in 2027 | Identity, architecture and volume |
| Customer five | Broadcom said shipments were strong | Shipments expected to more than double in 2027 | Identity, architecture and volume |
| Customer six | Count disclosed by Broadcom; program not identified in the cited coverage | Not stated | Name, design, volume and economics |
Broadcom describes these relationships as strategic and multiyear. That does not mean every customer uses the same XPU, supplier arrangement or volume commitment, and it does not make the projected revenue guaranteed.
Rank #2
- High-Performance AI Processing: The MX3 is designed to handle the most demanding AI computer vision workloads, delivering exceptional performance and efficiency.
- Flexible Integration: The MX3 can be easily integrated into your existing systems via its M.2 M-key form factor and support for Linux operating systems.
- Energy Efficient: The MX3 is designed to provide high performance while minimizing power consumption.
- Comprehensive Software Development Kit (SDK): The MX3 is supported by a comprehensive SDK that simplifies development and deployment.
- Hardware compatability: The MX3 is compatible with the PCI-SIG M.2 M-key 2280 Specification. It can be used with the Raspberry Pi 5 with a M-key 2280 HAT.
Why supply-chain commitments matter
Broadcom says it has secured capacity for critical inputs through 2028, including leading-edge wafers, high-bandwidth memory (HBM), substrates, T-glass and related substrate materials, advanced packaging and other constrained components. Chief financial officer Charlie Coz said customers share requirements two to four years ahead, giving Broadcom time to reserve capacity and sometimes help suppliers develop the needed technology. CRN and the transcript describe those commitments.
“Secured capacity” does not mean Broadcom owns all the factories or is insulated from execution risk. Yield problems, packaging bottlenecks, HBM allocation, substrate shortages, test capacity, geopolitical restrictions, supplier concentration, customer redesigns and cancellations can still delay shipments. Nvidia, AMD and other customers may compete for the same inputs.
Broadcom and Nvidia are not selling the same proposition
Nvidia’s role
Tan called Nvidia a formidable competitor and acknowledged that its chips continue to improve each generation. Nvidia’s broad accelerator platform, software ecosystem, developer adoption, networking products and rapid product cadence remain powerful alternatives for customers that want a ready-made system.
Broadcom’s role
Broadcom is positioning itself as the custom-silicon and networking partner behind hyperscaler-specific systems. Its pitch combines customer-specific design, SerDes and switching, advanced packaging, production scale and long-term supply planning. A custom accelerator may complement Nvidia in one workload and compete with it in another; Broadcom’s thesis is not that Nvidia disappears.
Rank #3
- ✅Powered by 26 Tera-Operations Per Second (TOPS) Hailo-8 AI Processor. 2.5W typical power consumption
- ✅Scalable, enabling simultaneous processing of multi-streams & multi-models
- ✅Enabling real-time, low latency and high-efficiency AI inferencing on the edge devices
- ✅Supports TensorFlow, TensorFlow Lite, ONNX, Keras, Pytorch frameworks
- ✅Supports Linux and Windows. Supports the temperature range of -40°C to 85°C
This model is most attractive to organizations with massive, stable demand. Smaller enterprises generally cannot justify commissioning and validating a bespoke accelerator and may be better served by commercial GPUs, cloud instances or managed AI services.
What the gigawatt figures do—and do not—tell investors
On the earnings call, an analyst calculated that Broadcom’s 2027 deployments could approach 10 gigawatts. Tan said that was the right way to think about the business, while cautioning that dollars per gigawatt vary substantially by customer. The near-10-GW figure is analyst math, not a standalone Broadcom forecast. Read the exchange in the transcript.
A gigawatt measures installed power capacity, not revenue. The same electrical capacity can contain different chip value depending on accelerator architecture, memory configuration, networking, rack density, cooling, deployment design and whether Broadcom supplies chips only or a broader rack-level package. There is no reliable public conversion from gigawatts to Broadcom revenue.
Financial reference points from fiscal Q1 2026
CRN’s summary of Broadcom’s reported fiscal first-quarter 2026 results gives the current baseline:
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Rank #4
- 48GB AI graphics accelerator
| Measure | Reported figure | Qualification |
|---|---|---|
| Total revenue | $19.3 billion | Fiscal Q1 2026; up 29% year over year |
| Semiconductor Solutions revenue | $12.5 billion | Fiscal Q1 2026 |
| Infrastructure Software revenue | $6.8 billion | Fiscal Q1 2026; up 1% year over year |
| AI revenue | $8.4 billion | Fiscal Q1 2026; up 106% year over year |
| Q2 total-revenue outlook | $22 billion | Fiscal Q2 2026 guidance |
| Q2 AI-revenue outlook | $10.7 billion | Fiscal Q2 2026 guidance |
| Net income | $7.3 billion | Fiscal Q1 2026; up 34% year over year |
These are fiscal-quarter figures, not calendar-quarter comparisons. The $100 billion statement is a 2027 management outlook and should not be added mechanically to the quarterly AI-revenue figures.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.VMware is the software counterpart
Broadcom presents VMware as a recurring-revenue and infrastructure-management counterweight to the more cyclical semiconductor business. The cited account reports VMware revenue up 13% year over year, total contract value booked above $9.2 billion, and annual recurring revenue growth of 19%. Broadcom forecast infrastructure-software revenue of approximately $7.2 billion in fiscal Q2, up 9% year over year.
Those figures are not interchangeable. VMware-specific growth of 13% is different from the entire Infrastructure Software Group’s $6.8 billion of revenue, which CRN reported up 1%. The $9.2 billion figure is bookings or total contract value, not recognized revenue.
Broadcom’s private-AI thesis
Broadcom has simplified VMware’s portfolio, moved customers toward subscriptions, emphasized VMware Cloud Foundation and promoted private-cloud infrastructure for workloads spanning CPUs and GPUs. Its stated strategy and early integration rationale are outlined in Broadcom’s VMware integration article and a company video.
Recommended Free Tools
Best Value
- Professional AI & Creator Workstation: AMD Radeon AI PRO R9700 GPU with 32GB GDDR6 is engineered for AI development, professional content creation, and compute-intensive workloads.
- Massive 32GB Memory Capacity: 32GB of GDDR6 memory on a 256-bit bus provides ample bandwidth for large AI models, 8K video editing, and complex 3D rendering.
- Advanced RDNA 4 with AI Accelerators: 64 Compute Units with 3rd Gen Ray Tracing and dedicated 2nd Gen AI Accelerators for groundbreaking AI performance and visual computing.
- Professional Blower Cooling: Efficient single blower design exhausts heat directly out of the chassis, ideal for multi-GPU workstation and server configurations.
- Enterprise-Grade Thermal Solution: Vapor chamber heatsink with industrial Honeywell PTM7950 thermal interface material ensures reliable cooling under sustained professional loads.
Tan’s argument is that generative and agentic AI will increase the need for private infrastructure, automation, security and workload management. Enterprises running sensitive or predictable AI workloads could use VMware as a common management layer across on-premises and cloud resources.
Why that outcome is not settled
- Some AI systems prefer bare metal or direct accelerator access, reducing the appeal of virtualization overhead.
- Subscription costs and portfolio changes may push customers toward Kubernetes-native stacks, public clouds or alternative platforms.
- Specialized AI services can avoid much of the infrastructure-management burden.
- Private AI may create software demand, but the scale and timing are not independently established.
“AI will create the need for more VMware” is therefore Broadcom’s strategic forecast, not an industry rule.
What could derail the $100 billion plan?
- Customer concentration: Six customers carry the central opportunity. A delay, redesign or cancellation at one can materially change the trajectory.
- Customer bargaining power: Hyperscalers can use multiple suppliers or develop more silicon internally.
- Nvidia execution: Better general-purpose accelerators could reduce the incentive to fund custom designs.
- Manufacturing risk: Working silicon still must achieve yield, packaging, testing and system-integration targets.
- Demand volatility: AI model economics, capital budgets or regulatory conditions could change deployment schedules.
- Revenue-definition risk: Rack spending and gigawatt estimates can overstate the portion that becomes Broadcom chip revenue.
- VMware adoption: Subscription acceptance and private-cloud demand may fall short of management’s expectations.
- Margin mix: Chip-only sales and rack-scale or system-level content can have different economics; Broadcom has not disclosed a future mix that permits a margin calculation.
Bottom line
Broadcom’s more-than-$100 billion 2027 AI-chip vision is a high-conviction management thesis built on real customer programs, custom accelerators, networking expertise and claimed component capacity through 2028. It is not a forecast that Broadcom will become a second Nvidia selling a universal GPU, nor proof that all projected deployments will become recognized revenue. The decisive questions are how much chip content each deployment contains, whether all six customers scale on schedule, and whether VMware becomes a preferred private-AI platform rather than simply another licensing option.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →
