Help me understand: why is 10GBase-T still so expensive?

So I've been pricing out a lab upgrade (currently running a mix of 1G stuff I got back in my day (which was 2019 (ancient history I know))) and the 10GBase-T NICs are just... stuck. $180-220 for a dual-port Intel X550-T2. Meanwhile I can get SFP+ cards for $40, DACs for $15. The gap makes no sense to me.

Actually I understand copper PHYs are complex (encoding, echo cancellation, the whole magjack situation) but we're years into 10GBase-T being "a thing." Shouldn't the chips be commodity by now? I remember when 1GBase-T dropped below $10 per port. What gives?

My use case: home lab, 10m runs max, already have Cat6a pulled (was optimistic (and maybe foolish)). Don't want to deal with optics if I can avoid it.

Anyone actually bought 10GBase-T recently and found a secret source? Or is this just the new normal?

SNAPPED UP a ConnectX-3 SFP+ for $12 on eBay last month

$12/YR ARE YOU KIDDING wait no it was one-time

10GBase-T still $200+ new, same as 2022. Price history flat. DACs dropped to $8 though.

My move: converted everything to SFP+. Even the switch. $80 used 24-port from RackNerd surplus. No regrets. 🏃

The PHY cost isn't the driver here.

For what it's worth, 10GBase-T requires significantly more DSP per port than SFP+ (which is essentially just a SERDES with minimal signal processing). The power draw alone is 3-4W per port versus under 1W for SFP+. That matters at scale, which suppresses volume, which keeps prices high.

The real correction: Intel essentially owns the consumer 10GBase-T market. Aquantia (now Marvell) tried to compete and got bought. No price pressure without competition.

SPF/DKIM/DMARC analogy: it's like trying to run a large mailing operation with only one reputable IP provider. Costs stay elevated.

My recommendation: use the Cat6a for 2.5G or 5GBase-T (much cheaper PHYs, Realtek and Intel both compete there) and run fiber or DAC for actual 10G needs.

• Power consumption: 10GBase-T PHY draws 3.5-4.5W typical, 6W+ at 100m. SFP+ DAC: <1W.
• Die size: DSP complexity means larger silicon per port. X550 is ~26mm² PHY area versus ~4mm² for typical SFP+ MAC/PHY combo.
• Volume: datacenter shift to SFP+ and now QSFP28/QSFP-DD reduced 10GBase-T demand by estimated 70% 2018-2024.
• Price elasticity: remaining market (SOHO, some enterprise edge) is price-insensitive.

Practical recommendation: your Cat6a supports 5GBase-T with $25 Realtek RTL8126 cards. For 10G, use SFP+ with 10Gtek DACs (model DAC-10G-CU3M, ~$12). I converted a 48-port lab in 2023. Total NIC cost: $340 versus $9,600 for 10GBase-T equivalent.

Fun little project: I had the exact same Cat6a situation (pulled it in 2021, very proud, very wrong about the future)

It actually works fine with 2.5G for now. But for 10G I just... got over the fiber thing? Bought a $25 Mellanox ConnectX-3, some €8 DACs from Hetzner's refurb store, and called it a day.

# my switch port config
interface ethernet 1/1/1
  no shutdown
  description "tom-nas"
  speed 10000
  duplex full

The DAC clicks in satisfyingly. That's worth something.

SFP+ runs cooler too. My closet dropped 4C average.

The RJ45 magnetics for 10GBase-T must handle bandwidth up to 500MHz with very strict return loss requirements. This is not a commodity component. The privacy angle: SFP+ modules with active firmware (certain brands from unknown sources) can in theory contain malicious logic. I am not aware of confirmed cases in homelab contexts, but to be precise, this is a consideration that 10GBase-T avoids by having no module-level intelligence. I converted my lab last year. Used DACs from interserver.net and Intel X520-DA2 cards. Total cost for 6 nodes: €340. The 10GBase-T equivalent would have been €1,800. The economics are, as far as I know, decisive.