SF Compute turns its pipeline into a supply signal on Lightfield
SF Compute provides GPU compute through clusters it operates for third-party providers and clusters it deploys itself. The company has to understand what each customer needs, then match that demand with capacity it can operate and support.
They provide a critical service in a market where demand outruns supply and securing the wrong capacity is one of the most expensive mistakes an infrastructure company can make.
Demand moved faster than the CRM could capture it
Inbound buyers arrive with highly-specific demands, including requested GPU model, GPU count, and networking requirements. On a busy day, as many as ten customers could arrive with ten different configurations.
Montana Showalter, SF Compute's Head of Go-To-Market, owns how the company responds to that demand. For her team, speed is critical when servicing inbound requests, because supply and demand are dynamic. Each request has to be understood, qualified, and matched against capacity from SF Compute or one of its providers. A slow intake process makes it harder to tell which requests the company can serve now and which recurring gaps should influence the capacity it sources next.
That urgency pushed the team toward logging requests in Google Sheets instead of their CRM. It was faster to type a customer’s requirements into a row than to create and update the corresponding records in Attio.
That solved the immediate data-entry problem but created another one. Customer emails, calls, relationships, provider information, and opportunity outcomes remained scattered across other systems. Montana and her team could record demand quickly, but they could not easily connect it to the full customer history or see what the market was repeatedly asking for.
A record fast enough that the team actually uses it
Montana brought SF Compute’s history from their previous CRM into Lightfield, then configured the pipeline around the information required to sell compute. Each opportunity automatically captures the GPU model, GPU count, managed services, start date, and other technical requirements that determine whether SF Compute can serve the customer. Lightfield extracts those details from customer interactions and populates them into opportunities with no manual data entry.
She uses chat to query demand requirements across the entire customer base, as well as to update records and find new business opportunities.
“One of the things that I love about Lightfield is getting to understand very quickly where we’re at with a certain customer,” Montana said.
That has changed how the GTM team works. A pipeline review no longer begins with a spreadsheet row stripped of its history. Each opportunity sits alongside the calls, emails, contacts, requirements, and prior decisions that explain it. Montana can update the record, analyze recent calls, or investigate a pattern in demand without first assembling the context from separate systems.
Extracting market signal from pipeline data
Once SF Compute’s opportunity history was connected, Montana could analyze demand across every deal rather than inspect opportunities one at a time. She started with a question that’s always top of mind for the business: how much customer demand had SF Compute lost because it lacked the right supply.
Lightfield gave her a full breakdown segmented by GPU model, GPU count, customer profile, and whether the workload was training or inference. All of this was grounded in the actual conversations her team had with prospects, not what somebody remembered to log.
The analysis Lightfield did has become an input into the company’s supply strategy. "Those breakdowns have been helpful for the hardware and supply team to understand where they need to target their attention," Montana explained. The reports show the supply team where unmet demand is concentrating across hardware configurations, customer segments, and workload types. Those patterns help the team decide which kinds of capacity and clusters deserve more attention, and whether repeat losses should change what it looks for next.
Making customer intelligence part of the company’s operating rhythm
The lost-demand analysis changed how Montana thinks about the information already sitting in SF Compute’s customer record.
The team is now building recurring reports to show which GPU models, cluster sizes, customer segments, and workloads are driving demand the company cannot serve. It is also exploring self-updating dashboards that show how those patterns change over time.
They’re also working to further automate how new demand enters the CRM. This includes automatic opportunity generation, as well as ingestion of Granola notes and Slack conversations into customer records. Together, these workflows will give SF Compute a more complete view of demand.
In total, this will give SF Compute a faster customer learning loop. Each conversation adds context to the record and increases their understanding of what their customers need. Now, they can respond to each request with more context while learning from the demand it cannot yet meet. Every interaction makes the next response better informed and helps SF Compute improve what they offer their customers.