Nvidia has now turned its investment jets on a mobile chip firm, Taiwan-based MediaTek. The AI chip giant has invested $3.5 billion in MediaTek this week.

This investment — in the form of convertible bonds that will be issued by MediaTek — is even bigger than the $1 billion that Nvidia poured into Nokia back in October 2025. That infusion inspired Nokia to focus on Nvidia GPU-based AI-RAN, which it expects to start trialling this fall.

As for Nvidia, the chip giant has initiated more than $600 billion in total financial frameworks and direct credit support, all in the name of AI infrastructure buildout, just in the month of August alone.

So how might this latest Nvidia investment impact MediaTek’s future?

On the face of it, MediaTek and Nvidia are collaborating in several major areas: AI supercomputers, enterprise, computer chips and automotive platforms. MediaTek will integrate Nvidia's NVLink Fusion and NVHBM interconnect technologies to design custom processing units (XPUs) for data centers.

Previously, MediaTek had been trying to expand its presence in the custom data center ASIC chips sector, hoping to generate $2 billion in revenue from that business in 2026. This deal will allow MediaTek to continue to build custom chips designed by AI labs and cloud companies, while using Nvidia’s NVLink Fusion platform.

The MediaTek deal should help Nvidia maintain its lead as the data center framework provider. The chip company has been losing a little GPU ground as AI and cloud companies, including Anthropic, Google and Microsoft, have been investing to build their own chips to break away from Nvidia’s AI stranglehold.

Interestingly, the deal could also have some affect on MediaTek’s limited arrangement with Ericsson, which has been using MediaTek for network optimization and 5G and 6G prototypes. Ericsson has worked on AI software with Nvidia for T-Mobile before. We’ll see if the Nvidia AI links come into play as Ericsson moves into 6G towards the end of the decade.

Researching accounts. Building lists. Writing sequences.

There's a better use of your team's time.

Apollo is the AI revenue engine that handles the busywork, so you can stay focused on selling.

Plus, everything you need is in one place:

Stop doing busywork and start building pipeline, faster.

With Apollo — the AI revenue engine powering 4M+ users.

Sign up for free

I know you’ve all heard about Elon Musk and SpaceX’s grandiose plans to grab gobs of customers from the Big 3 mobile operators and eat into their combined annual revenue of around $600 billion. 

But let us go into what Musk and friends would have to do to actually make the plan work.

SpaceX and Starlink Mobile have said they will deliver a service by bolting tiny base stations, called femtocells, onto existing Starlink satellite dish terminals to provide a service that is “better” — according to the most rich snake oil salesman in the Musk — than the major mobile operators. SpaceX President Gwynne Shotwell said that the femto service would reduce the massive capital needed to put up and support a nationwide 5G network.

Femtocells typically have a short coverage range of 10 to 30 meters and are typically used for in-home cellular coverage extension. Femtocells are the lowest range type of small cell.

Analysts have described the plan as nuts and said it won’t have great coverage or be able to compete with existing services from AT&T, T-Mobile and Verizon. There’s a small subset of analysts and commentators who think that SpaceX wants to jump ahead of the  mobile operators and take the AI token opportunity away from them. 

The company, however, has never said anything at all about AI or tokens in relation to their mobile femtocell plan. It only mentioned mobile consumers and Big 3 operators in that part of the earnings call.

I should note that SpaceX doesn’t have a press department. So, it is difficult to get any further information out of the organization beyond what the executives have said on their earnings call.

It’s the spectrum, stupid

SpaceX, of course, has lauded the 65 MHz of spectrum it bought from Charlie Ergen’s EchoStar. Musk’s satellite mavens said that combining the 65 MHz EchoStar spectrum with a planned tenfold increase in next-generation mobile satellites, which are expected to launch in mid-2027, will make Starlink Mobile 100 times better than current operations.

The EchoStar spectrum that SpaceX bought consists of blocks of AWS-4 spectrum, H-Block frequencies and a little tranche of AWS-4 radio frequency. This provides the company with nationwide coverage in the U.S., although with gaps in its spectrum scope.

“Most of that stuff is kind of shit,”  said Earl Lum, president of EJL Wireless, of the spectrum bluntly,  on a call with me last week. He expects that Starlink will have to bid on the Upper C-Band (3.98-4.14 GHz) auction that the FCC plans to have happen by July 2027, in order to have any hope of getting their femto scheme off the ground.

Using Dish’s defunct network

“I don't see how SpaceX can operate a network without some macro sites,” noted Recon Analytics analyst Daryl Schoolar. “The company needs network coverage and small cells on subscriber houses won't come close to providing that needed coverage. Small cells are better for adding capacity and spot coverage than they are for full geographic coverage. “

One way that SpaceX and Starlink Mobile could add some macro sites in a timely manner, some commentators have suggested, is the now-defunct Dish network, which operated on the same spectrum that SpaceX now owns.

As Lum said, however, the tower companies — American Towers, Crown Castle and SBA — that are currently pushing for around $7.6 billion in disputed lease claims with Dish Wireless would have to have “renegotiated leases” if SpaceX is to take over the Dish network.

EchoStar Chairman “Charlie [Ergen] stiffed everyone and didn’t pay out on those leases,” Lum said.

As part of its bankruptcy, there’s an government mandated escrow account that Dish has to pay into to make everyone whole, but at the moment the tower companies are still waiting for their money, the analyst noted.

“Once he does pay up and everyone’s good then SpaceX could renegotiate the leases on all of those towers,” Lum said. All of this, however, would take time, which is a concern for SpaceX, as executives have said that the town-and-city-capable service will be able late in 2027. 

T-Mobile 2.5 GHz licenses?

Jim McKenna, CEO at RedZone Wireless suggested on LinkedIn recently that SpaceX could potentially grab some 2.5 GHz licenses for a nationwide network.

”The conventional wisdom is that T-Mobile has the 2.5 GHz band locked up … T-Mobile controls approximately 92% of the band, yet owns only about 38%,” McKenna said.

He suggested that SpaceX — or another party — could lease some or all of the 54% of largely educational 2.5 GHz licenses that T-Mobile leases to form their own nationwide network. On paper, this looks like an attractive proposition.

Having covered the 2.5 GHz license game for a while, however, I don’t think it will be that easy. Sprint’s - remember them! - holdings of 2.5 GHz licenses were a big part of the reason T-Mobile paid $26 billion for that operator in the first place.

Given that T-Mobile’s 5G layer cake strategy is built on 2.5 GHz spectrum, I very much doubt that the operator will allow its cellular advantage to be swiped out from under it without a fight.

“All of those were long-term contracts,” said Lum of the T-Mobile 2.5 GHz leases. “Depending on when those [leases] end, that’ll be a bidding war, which could be bad for T-Mo,” he said.

That’s not something that will occur any time soon though. “It’s not going to happen tomorrow,” Lum stated.

This still leaves D2D provider SpaceX with a big terrestrial problem.

“They can’t really make their supposed network with just small cells,” he stated. “At this point, I think they should have just said nothing...and thought about what is their strategy. Because without more spectrum, what is the point?” 

The difference between growing stores and stagnant ones isn't more effort. It's better insights. StoreClaw analyzes your Shopify and Amazon data, surfaces your biggest growth opportunities, and helps you increase GMV while protecting profit. Start free with bonus tokens. No credit card required.

Try it FREE today

Starlink Mobile’s ability to build or lease a terrestrial network remains very much an open question, but the Starlink Broadband sector of SpaceX’s telecom business is actually growing in rural areas, as Ookla noted.

The firm’s latest research, however, also illustrates that Big 3 operator-backed FWA is growing steadily in both town and country.

Ookla just just published last week its latest report on how U.S. 5G fixed wireless access (FWA) connections from Verizon, T-Mobile, and AT&T topped 17 million in Q1 2026, capturing 13.74% of the nation’s 123.75 million home broadband subscriptions. Ookla editorial director Sue Marek told us that operator FWA connections are growing in urban as well as rural areas.

For each of the Big 3 U.S. carriers, FWA download speeds decreased in the second quarter of this year versus Q1, per Ookla.

Meanwhile, upload speeds decreased across the board, as well.

“The three national providers did see their speeds dip in the second quarter of 2026 compared to the first quarter of 2026 and we think this is a seasonal phenomenon due to the impact of foliage and dense leaves on trees,” Marek said. “This tends to occur in the spring and summer months.”

In fact, “leaf obstruction is a problem that RF engineers deal with on a daily basis,” Marek added.

With more urban users moving to budget-friendly FWA home broadband services, will speeds dip even further as limited spectrum gets spread thinner and thinner?

“That’s certainly possible. But so far we haven’t seen FWA speeds decline and stay that way,” Marek said. “It appears that so far FWA providers have been able to engineer their networks appropriately to  accommodate the number of subscribers on their networks.

“However, this is the first time we’ve used the U.S. Census Bureau’s urban-rural classifications to categorize our Speedtest Intelligence data, so I can’t really speak to whether the urban speeds alone are declining. But we’ll continue to track this trend in future reports,” she added. 

Creators lock in holiday content calendars 90 days out, before brands figure out commissions. Waiting too long to launch an affiliate program means less runway to build demand and a missed shot at the best partnerships.

The 90-Day Holiday Sprint covers commissions, recruiting, and scaling a program at Day 30, 60, and 90.

Download the Free Guide

Meanwhile, data released by Ookla in April covering Q1 shows that FWA speeds across the U.S. were higher than those offered by Starlink Broadband.

“It’s difficult to say that performance alone is going to provide Starlink with an opportunity to lure FWA customers away from these three FWA providers,” Marek said.

“However, I do think Starlink is a competitive threat to many types of broadband providers, not just FWA. Starlink’s performance is steadily improving and it is aggressively promoting its services and offering discounts on its equipment. In addition, it appears to be targeting customers where it has excess capacity such as rural areas.”

Starlink’s rural competition is often DSL or cable services that are limited by the old copper lines they run on, or slightly older satellite options like Viasat or Hughesnet.

Next TMT Networks has previously reported on how analyst Craig Moffett sees Starlink Broadband and Mobile as a nice global business which should have around 91 million customers worldwide by 2031